
- Blog Posts


- Blog Posts

- Blog Posts
All Categories
SEO
SEA
Social Media
Web Analytics
Growth Marketing
Artificial Intelligence
Success Stories
Inside internetwarriors
Search engine optimization with Bing SEO
Oct 16, 2019

Thorsten
Abrahamczik
Category:
Search Engine Optimization

As a website operator, you want your site to be easily found. You often think of Google, but not of the search engine Bing, right? For many website operators, search engine optimization is an abstract topic, and Bing SEO often gets pushed into the background. This is probably because Germany is a Google country. According to Statcounter, Google had a market share of 94.5% in Germany in 2019, while Bing only had 3%. [caption id="attachment_25213" align="aligncenter" width="1024"] Fig. 1 Statcounter statistics on search engine usage in Germany in 2019[/caption] What features does Bing offer? With Bing's previously unheard-of preview function in search engines, users receive useful information about content and internal links even before they enter the website. Here, the displayed content is actually drawn from the page content and not, as with a snippet, from a meta-description. [caption id="attachment_25221" align="aligncenter" width="1024"] Fig. 1: Statcounter statistics on search engine usage in Germany in 2019[/caption] Still, you shouldn't ignore the Bing search engine. Users who come via Bing often have better user behavior than those from Google. They typically view more pages, stay longer on the domain, and have a higher conversion rate. It may therefore be worthwhile for you as a provider to engage this audience. Also, the results of Apple’s Siri service and Amazon’s Alexa are based on Bing. [caption id="attachment_25223" align="aligncenter" width="1024"] Fig. 2: Comparison of user behavior on Bing with Google[/caption] Commonalities between Bing and Google In principle, you are already optimizing your website for Bing if you carry out search engine optimization according to Google guidelines. This is partly because Bing uses similar criteria. It is also because some SEO standards from Bing, Google, and other search engines were created in alliance together. An example of this is the structured data from schema.org . Further aids for Bing search engine optimization A look at Bing's search results page is also very helpful. Bing specifies much more precisely where the data comes from. In addition to pure search engine optimization, revising other profiles on the Internet such as Xing, Wikipedia, etc., can also lead to better visibility on Bing. The optimization of multimedia elements on the website can also show better effects than on Google. This is visible with Bing image search or video results. By strategically using structured data and consistent use of alt titles, you can position images and videos very well in the search engine. Furthermore, Bing search offers some options that are not available in Google. For example, individual elements in images can be selected for search. Videos are supplemented in the results with a lot of information, such as whether it is a trailer. Also, videos can already be watched in preview, and Bing considers the search query history to display the most suitable results. [caption id="attachment_25225" align="aligncenter" width="1024"] Fig. 3: Identification of elements in Bing image search[/caption] What to pay attention to Microsoft offers various guidelines and tools for its search engine to support users in achieving the best possible search engine optimization. The following three are the most relevant from our perspective: Bing Webmaster Tools Submit sitemap.xml Block URLs Monitor reports Control crawls Test mobile-friendliness We warmly recommend using the Bing Webmaster Tools to gain better control options. Of course, the domain can also rank in the Bing search engine even if it has not been registered there. Nonetheless, it is interesting to see how Bing assesses the website compared to Google. Bing Webmaster Guidelines Bing offers the very informative Bing Webmaster Guidelines on its websites. These provide users with initial help on the topic of search engine optimization. The instructions are thematically subdivided and clearly presented. They are helpful for both experienced users and newcomers to the topic of Bing SEO who want to know what the search engine operator values. Markup Validator If you use structured data on your website, you can check the correct implementation with the Bing Markup Validator . Supported formats are: HTML Microdata Microformats RDFa schema.org OpenGraph from Facebook Please note, however, that you must register to use the tool with the Bing Webmaster Tools. This also applies to the SEO Analyzer and the keyword research tool. Peculiarities of the Bing search engine We generally want to give you three tips that will help you with Bing SEO: 1. Exact Match URLs: Bing places much more emphasis on the correct name in the domain than Google does. Therefore, if you own a domain that includes your brand name, it will automatically rank better on Bing. 2. JavaScript Currently, Bing is not as advanced as Google in accounting for JavaScript on websites (as of 10.2019). This means that while your site may rank well on Google, it might be placed lower on Bing. Sites that dynamically load content through JavaScript are particularly affected. However, if your site uses minimal JavaScript or does not load content dynamically, you can overlook this point. 3. Google Adjustments Google frequently makes adjustments to its SEO guidelines, suddenly weighting certain features more or less. The last known adjustment was the removal of "noindex" from the robots.txt file. It's important to note with these changes that they apply only to Google and not automatically to Bing. So don't be alarmed by them. Conclusion The Bing search engine offers many unique features to differentiate itself from its bigger rival, Google. At the same time, its users exhibit consistently better behavior on websites, regardless of the industry. To rank well here, you only need to make a few specific adjustments, but you can use Bing's excellent tools to closely monitor your developments. How can we support you? Do you want to increase your conversions and revenue through organic search? Are you already ranking well on Google and now want to kick off with Bing? We support you in Bing SEO and thus achieve higher rankings in the Bing search engine. Contact us here , we look forward to your inquiry!
On-page SEO Strategies for Better Search Engine Optimization
Jun 18, 2019

Thorsten
Abrahamczik
Category:
Search Engine Optimization

SEO is a commonly used term among online marketing managers and often comes in conjunction with off-page optimization . Both terms describe the search engine optimization of your own website . The aim of targeted SEO optimization is to improve the ranking of your own website in search engines, generate more traffic, and ultimately lead to more conversions or leads. However, marketing managers usually only know a portion of the actions that can be performed. In this article, we explain what on-page SEO is all about and offer initial tips for practical implementation . On the internet, you will come across various spellings, but it always means the same thing. Possible terms are: Onpage SEO, Onpage optimization, On-page optimization, On Page optimization, SEO on-page, SEO optimization, or even on-site optimization. On-page Analysis – The first step in On-page SEO At the beginning of your work, you should perform an onsite analysis of the entire website. First, you analyze the technical SEO to identify possible technical pitfalls that could affect the crawling of the website. In the next step, you examine the contents of the website in order to carry out content optimization. After investigating both areas, you can derive a prioritized action plan that you implement step by step. To track the development of your website in SEO, you can use Google's Google Search Console. This serves as an interface between website operators and Google, showing which URLs rank well, for which keywords the domain can be found, and where there are problems on the website. It also visually represents how often individual pages are displayed on Google's search result pages. Technical SEO – The Foundation of On-page Optimization Before you can start optimizing your content, your technical SEO must be implemented flawlessly. This ensures that the website can be crawled easily by search engines and that all content is read and processed accurately . If technical SEO is not implemented correctly, it may happen that content optimizations, the so-called content optimization, do not take effect properly because the search engines do not have access to the content. Therefore, technical SEO is the foundation of every SEO on-page optimization. Crawling – Can the search engine reach your website? Crawling solely refers to accessing content and has nothing to do with indexing. Search engines like Google certainly crawl pages, but they do not index all of them. This can occur for various reasons, including poor search engine optimization. To improve the relevance of content, only the content that needs to be processed by search engines should be made crawlable. Pages that are not important, such as search results pages or the imprint, should be excluded from crawling. This is referred to as optimizing the so-called crawl budget. It determines how many pages of a website can be crawled and is set individually by search engines for each domain. In this context, there is an important distinction: Crawling is not equivalent to indexing! To control crawling, use the meta tag "robots" and the "robots.txt" file . However, it should be noted that the instructions contained in the "robots.txt" are merely a suggestion to search engines and can be ignored entirely. Furthermore, search engines can also reach excluded pages through other ways, such as backlinks. Therefore, using the "Disallow" function is not a reliable method to exclude pages from crawling. On the contrary, incorrect use of the robots.txt can lead to major problems on search results pages, as Figure 1 illustrates. Figure 1: Websites indexed by Google for which no descriptions can be displayed because they are excluded from crawling in the robots.txt. In May 2019, Google updated its in-house Google Crawler to the current Chromium version 74. This is an important note because its predecessor was outdated and supported only few modern web technologies. The new crawler can now recognize modern SEO optimization and perform better on-page analyses. SUBSCRIBE TO NEWSLETTER NOW URL Structure, Internal Linking, and sitemap.xml are important SEO On-page Factors Next, you need to review the URL structure of your website. Key questions are: Is it readable and clearly understandable for humans? Can users tell where they are on the website? Is it not too long? The URL structure should not exceed 5 hierarchy levels . This ties in with the structure of the website. Today's on-page SEO is not about optimization for search engines, but for users. This means that content must be easily and quickly accessible . Excessive cascading of individual webpages is therefore not advisable, as it only leads to more hierarchies. John Mueller, Webmaster Trends Analyst at Google, announced on 03/05/2019 via a Webmaster Hangout that the internal linking of pages should be weighted even more than the URL structure. Visible URLs are primarily relevant for user experience. Internal linking of pages should instead be done in a way that is relevant to the topic, to achieve good on-page optimization. In this context, it is also referred to as siloing. For good internal linking, you must ensure that links are made only within a topic (silo). For example, all articles on web analytics should link to each other, but not to articles on SEO. Therefore, you should always develop a linking concept and check, especially with links that have JavaScript functionality, whether crawlers can find and follow the links. If you have similar content, you should use the canonical tag. Imagine you run an online shop and offer a t-shirt in five colors. You provide a separate page with a unique URL for each of these color variants. In this case, you would have duplicate content on the domain because all pages would be textually identical and would only differ through the color specification. With the canonical tag, you can indicate on all pages which of the five URLs is actually relevant to search engines and which serve solely as added value for the user. Additionally, you should provide search engines with a sitemap.xml file , a file listing all the URLs that are to be indexed. Therefore, this file should not include URLs set to "noindex" or those excluded from crawling or indexing in any other way. Avoid Duplicate Content with SEO On-page The topic of duplicate content or duplicate content is one of the main focuses of SEO on-page optimization. Often, the same pages are accessible under multiple URLs. Classic examples are: URL with http and https URL with www and without www URL with trailing slash and without URL with very similar content For this reason, it is essential to set up redirects and communicate very precisely to search engines via canonical tags and sitemap.xml which websites should actually be indexed. Imagine you run an online shop and offer a t-shirt in five colors. You create a separate page with a unique URL for each of these color variants. In this case, you have duplicate content on the domain because all pages are textually identical and differ only in color indication. With the canonical tag, you can indicate on all pages which of the five URLs is actually relevant for search engines and which pages serve solely as added value for the user. During a redirect, users calling a URL with http are automatically redirected to the https variant. This happens so quickly that they often do not even notice it. Pagespeed – How Fast Does Your Website Load? For several years now, Google has recorded more accesses via mobile devices than via desktop devices. It is therefore important to offer a fast-loading website . This can be implemented in several ways: Ensure small file sizes and short source codes. This optimizes the website as a whole and ensures a good user experience. Optimize visible content through prioritized delivery of source code . With this measure, users see the first content in the visible area long before the entire website is loaded. This improves the perceived load time. Switch the Hypertext Transfer Protocol (http) to version 2 (http/2). This is optimized for mobile devices and allows parallel loading of various files, as well as preloading of content. We also recommend using https for good on-page SEO. Figure 2 illustrates the Google Page Speed Test, which examines exactly these topics: Figure 2: With the Google Page Speed Test, you can see which files are too large and which files delay optimized delivery of visible content. Mobile Optimization is Becoming Increasingly Important In addition to the load time, the display of the website on mobile devices must also be ensured. This is often achieved with responsive design, which ensures that the website adapts automatically to the screen size of mobile devices. If you are planning a relaunch soon, this issue must be addressed in your concept. Structured Data for Good Onsite Marketing With structured data, individual content is specifically tagged for search engines. This may include company information, product information, recipes, events, or, more recently, FAQs. There are many templates for using structured data on a website, although only relatively few are supported in Google optimization. The benefit is that Google can better understand the content and display it separately on the search results page. Further SEO On-page Optimizations The technical SEO area also includes further measures such as Progressive Web Apps, image tagging, or multilingualism. These are the "fine-tuning" aspects in the field of technical SEO optimization. Therefore, these detailed topics are not elaborated here. Content Optimization – The Second Step in On-page SEO Once the website has been improved with technical on-page SEO measures, you can now start optimizing your content. This is basically divided into two areas: Meta Information: Optimizing Page Title and Meta Description The optimization of the page title and meta description is referred to in on-page SEO as the optimization of meta information. The page title as an SEO criterion is particularly crucial as it should contain the keyword and the brand. At the same time, it should not be too long to remain readable. You always see the page title in the tab of your browser as well as on Google's search results page. The meta description itself is not SEO relevant but still has a significant indirect impact on onsite marketing. The meta description is displayed on search results pages of search engines if relevant. By directly addressing users with a call to action, you can improve the click-through rate on your results and thus achieve a better ranking. Figure 3: Google search results with the page title in blue and the meta description in black. Content on the Page Itself – What Can the User Expect? The actual content must meet user expectations, otherwise, it will lead to high bounce rates. Remember that content should be created for users in the context of on-page optimization, not for search engines. For this reason, Google places great emphasis on how readable the content is. Furthermore, the texts must be well-structured and organized with headings. Increasing user interaction with the website is also desirable. This can be achieved through videos, images, image galleries, comments, or similar features. It is proven that longer dwell times are associated with an increase in conversions. If the content is well created, Google may use it as a Featured Snippet. This is the position 0 on the search results page, where Google directly answers the user's question on the search results page. Marketers have no influence over the use of a Featured Snippet and cannot predict when a snippet will be displayed and when not. Getting your own content into a Featured Snippet is therefore considered the pinnacle of Google optimization for content. Figure 4: Representation of a Featured Snippet on Google's search results page. Google Jobs – A Brand New Feature With Google Jobs, the search engine giant introduced a brand new feature in Germany in June 2019, which will lead to many on-page SEO optimizations in 2019. When users search for a job title, they are shown a list of available job offers. After clicking, they are directly redirected to the company's page, where they can apply for the job in the next step. However, to do this, website operators must use structured data and provide very specific content on the website. What We Can Do for You If you want to improve your positions in search engine rankings and thereby increase your number of conversions, we offer comprehensive support in the area of on-page optimization. Our on-page SEO measures are coordinated with other online marketing measures. Feel free to contact us, we look forward to your inquiry.
Opt-In, Initial Insights from Practice
Jun 28, 2018

Thorsten
Abrahamczik
Category:
Web analytics

Opt-In – Impact on Online Marketing through the EU Cookie Directive Under the new General Data Protection Regulation (GDPR), many marketers have experienced significant confusion regarding the EU Cookie Directive and the Opt-In and Opt-Out procedures. Additionally, there is uncertainty about the e-Privacy regulation, which is expected to become mandatory in 2019. In our article "No Google Analytics without Google Analytics Opt-Out Cookie" , we have already discussed the necessity of a Google Analytics Opt-Out Cookie on the privacy page. In this article, we want to explain the Opt-In procedure, which requires the explicit consent of the user for analysis and marketing measures. We will also illustrate how this procedure affects all online marketing channels and activities. What is the Opt-In procedure? The Opt-In procedure is based on the increasingly popular cookie notice, which mentions the use of cookies on websites. As shown in images 1 and 2, the formulations were revised on May 25, 2018, and supplemented with additional information on the use of cookies in many cases. Fig. 1 Old cookie notice as it was used on https://www.internetwarriors.de before GDPR. Fig.2 Current notice, allowing users to exclude themselves from tracking. Since this revision, many users have had the opportunity to agree to or decline the use of cookies on individual websites. Once users make a decision here, the use of cookies, aside from specific exceptions like session cookies, must be respected across the entire domain. However, many lawyers and data protection officers interpret the GDPR differently, resulting in users being offered various solutions. These range from simple cookie banners without selection options to Opt-In procedures. The impacts of the Opt-Out procedure are already known. However, with the Opt-In procedure, only a very few companies have experience. For this reason, we tested the Opt-In procedure within the framework of GDPR cookies to gather initial insights that we can consider in future implementations. Distinction between Opt-In and Double Opt-In Before we begin with the implementation and the impact on traffic, we need to differentiate between Opt-In and Double Opt-In: Opt-In: An information banner is displayed to the user on accessing the website, informing them about the use of cookies and, if necessary, their purposes. The user must also explicitly consent to the use of cookies before web analysis and marketing measures may be carried out. If they do not, neither tool may be used. Double Opt-In: This procedure is primarily used in email marketing. Upon subscribing to a newsletter, the user receives a confirmation email, requiring them to actively confirm their subscription. As you can see, both procedures are independent of each other and have nothing in common. Changes in traffic due to the implementation of Opt-In As part of our Opt-In investigation, we examined the traffic development on 10 websites in Google Analytics before and after implementing the Opt-In. The Google Analytics screenshot in image 3 shows the number of sessions of a website on a daily basis, before and after the implementation of Opt-In. Fig. 3 Traffic development from May 24, 2018, to June 21, 2018. Opt-In was implemented on June 8, 2018. Comparing the period after implementation with the period before implementation and excluding the day of implementation, the following traffic changes arise: Fig. 4 Comparison of developments in Google Analytics in the periods June 9, 2018 – June 15, 2018, and June 1, 2018 – June 7, 2018 Other websites with about 5,000 sessions per day even show deviations of 83% - 85%. Only a few websites have a smaller deviation than shown in the screenshots here. Configuring Opt-In with a Step-by-Step Guide To help you understand how the entire procedure works, we would like to give you a detailed Step-by-Step explanation. Additionally, at the end of the article, we offer you the chance to download our configuration of the Google Tag Manager container so that you can import it into your Google Tag Manager and gain experience with the implementation. GDPR Cookie Notice on the Website An essential requirement is a cookie banner on your website, informing users about the use of cookies and giving them the option to activate or leave analysis and advertisement cookies deactivated. For simplicity's sake, we use the popular solution Consent by Insites for our attempt. We have also integrated this script on our website. Via the Download menu, you can configure a banner that you only need to copy into the source code of your website afterward. During this process, you have to decide whether you want to use the Opt-In or Opt-Out procedure. In our current scenario, we use the cookie notice for Opt-In. Subsequently, the banner is displayed immediately. Thus, implementation is very easy to carry out even for less technically skilled individuals. Storing User Decision in a Cookie Once you've embedded the banner on your website, it will be displayed to all users. However, initially, nothing else happens, no cookies are blocked yet. Insites itself uses a cookie named "cookieconsent_status" to store the user's decision and not display the banner again on their next visit. This decision is valid for one year. The cookie values can be seen in image 5: "allow" for consent "dismiss" for rejection You also get the expiration date of the cookie, from which the browser no longer considers the cookie. We can read the "allow" and "dismiss" values with the Google Tag Manager and take them into account for triggering Google Analytics, Google AdWords, Affiliate, etc. The decision of an Opt-In should not be limited to web analytics using Google Analytics, Etracker, Webtrekk, etc. alone. All remarketing and conversion tracking from other providers should also be considered. Fig. 5: Status of the "cookieconsent_status" cookie for Opt-In Consideration of Do Not Track After considering the GDPR cookie decision by the user, we want to consider a second option of rejecting analysis and marketing cookies. This involves the Do Not Track procedure . In this case, the browser sends information to the server with each new page view that no user profile should be created and personal activities should not be tracked. Image 6 shows the setting in Firefox's "Privacy & Security" section. Fig. 6: Activation of the "Do Not Track" information in Firefox's privacy settings Do Not Track is integrated into all relevant browsers like Google Chrome, Mozilla Firefox, Apple Safari, etc., but is disabled by default. Therefore, the user must make a conscious decision and manually enable Do Not Track. If they do, website operators should respect this decision if they offer the Opt-In procedure. Interaction of individual configurations in the Google Tag Manager To configure Opt-In in the Google Tag Manager and consider GDPR relevant cookies, we have defined the following rules: Has the user explicitly agreed to the use of cookies for Opt-In? If yes, we check whether the user has activated Do Not Track If no, we keep all tracking disabled Has the user activated Do Not Track If yes, we keep all tracking disabled. This rule also overrides the previous rule if the user has agreed to tracking on the banner If no, we check whether the user has consented to the use of cookies. Only if both conditions are fulfilled will analysis and marketing cookies be activated The user has consented to the use of cookies The user has deactivated Do Not Track If even one value deviates, the cookies remain blocked. This way, the website operator offers maximum protection for users from cookie capture. At the beginning of the article, we showed that this setting in the Google Tag Manager resulted in significant traffic loss in Google Analytics. But since all remarketing and conversion tracking is also blocked, website operators can no longer tag their users and can measure success significantly less. Measuring Do Not Track Usage on Another Website Currently, according to our non-legally binding understanding, there is no obligation to use Opt-In tracking. However, this may change with the e-Privacy regulation in 2019. Regardless, it is not known to us that the Do Not Track feature is a mandatory measure for website operators. For this reason, we analyzed the use of Do Not Track on an eleventh site. This site serves family entertainment and is characterized by a high national as well as international traffic. It also serves both genders and age groups from infants to great-grandparents. We consider these numbers a good cross-section of society. In image 7, we have juxtaposed the number of sessions and accesses with activated Do Not Track. For measuring activated Do Not Track, we use "Unique Events" in Google Analytics, as this value is "session-based" and thus provides a comparable data basis. Fig. 7: At 10% of all sessions, Do Not Track is activated in the period June 13, 2018 – June 20, 2018 The collection period is June 13, 2018 – June 20, 2018. It is clearly visible that in 10% of all sessions, Do Not Track is activated. Here, users have made a very conscious decision not to be tracked. Learnings from the Test These are very valuable and important insights for us. The Opt-In procedure significantly reduces the metrics in the analysis and marketing tools and makes it considerably more difficult to capture users. If the use of Opt-In becomes mandatory, other methods would need to be developed to continue offering online marketing in the same quality. If you, as a website operator, need to decide between Opt-In or Opt-Out, you now know the pros and cons. We are also happy to offer our Google Tag Manager container configuration for download. Fill out the following form, and we will send you the download link by email. The .zip file can be easily opened, and you will find a .json file inside. When you are in your Google Tag Manager, click on "Admin" and then on "Import Container." Subsequently, select the .json file and import it into your Google Tag Manager container. You will then find all the templates we created. It will be exciting to see how the e-Privacy regulation impacts the EU Cookie Directive and how users take advantage of Opt-In options. What can we do for you? If you are unsure whether you need Opt-In tracking or if you experience difficulties implementing the Opt-In or Opt-Out procedures, we are here to help you. We support you in the implementation of your online marketing strategies and can quickly make technical adjustments to your website, should the new e-Privacy regulation require it.
Early configuration of Google Tag Manager with the Tag Manager Injector
Aug 31, 2017

Thorsten
Abrahamczik
Category:
Web analytics

The Tag Manager Injector, as a Chrome browser plugin, allows for easy setup of the Google Tag Manager within Google Chrome. Without long waiting times, we can start servicing our clients and are not forced to wait. Meanwhile, the client's IT can take its time integrating the Tag Manager code into the website's source code. At the beginning of our collaboration, we work closely with our clients to create tailored tracking concepts. At this point, it is often not yet clear what exactly needs to be tracked and which metrics provide value to the client. Once the tracking concept is finalized and approved by the client, we begin implementing it. First, the Google Tag Manager code needs to be embedded into the site's source code. However, this can typically take several days, as the client's IT may not be able to implement it immediately. To start configuring in the meantime, we use the Chrome browser plugin Tag Manager Injector. This makes us much more independent from the client's IT and allows us to work faster. Requirements for Using the Tag Manager Injector To use the Tag Manager Injector, a Google Tag Manager account must first be created. Within this account, a container must be created. The Google Tag Manager then provides the code for the container, which the IT must integrate. At this point, we can already start using the Tag Manager Injector to configure the newly created container. In this context, the unique container ID, which identifies this particular container, is important. Fig. 1: The container created in Google Tag Manager with the container ID Using the Tag Manager Injector After this process is completed, you go into the newly created container and create the first tags, triggers, and variables. At this point, the container can already be configured as it is intended to be used later. There are no limitations here, as third-party tags and scripts can be used. After setting up the initial tags, Google Tag Manager's preview mode should be activated. This way, you can check whether the tags have been configured correctly. The next step is to install the plugin. Once the plugin is active, it can be used. To do this, click on the plugin itself and you will see a simple input screen where you enter the container ID of the previously created container. Next, under the point “Include Domain(s)”, enter the domain where the Google Tag Manager container should be used. Then, just click the “Start” button. Once the Tag Manager Injector is active, the area at the GTM container ID turns green. Fig. 2: The configured and usable Tag Manager Injector By activating preview mode in the container, the regular Google Tag Manager preview window opens on the website in the browser. It's now easy to see which tags fire, what variable values are displayed, and what the data layer looks like. If you want to push specific values to the data layer, the Tag Manager Injector offers a clearly visible input field “Push to the data Layer”. You just need to enter the respective information there, and the data will be transmitted. Fig. 3: The Google Tag Manager preview window in the browser In the real-time report of Google Analytics, initial accesses can now be seen, showing that Google Analytics tracking is working. How Can We Help You? Would you like to implement web analytics on your page according to a defined tracking concept but are unsure how to do this exactly? Do you want to measure and increase conversions more effectively but have problems implementing additional tracking? Contact us and we will gladly help you improve your web analytics.
Record entries of domain management and what you need to consider when making adjustments
Jun 22, 2017

Thorsten
Abrahamczik
Category:
Search Engine Optimization

As a website operator, you must also deal with the management of your domain. This involves not just purchasing a domain but also configuring it for web servers and email servers. This primarily concerns larger companies or agencies that manage domains on separate servers, which are operated by the actual web or email servers. An example of such a scenario includes providers like NICdirect or 1Blu, where domains are purchased and managed, with a specialization in hosting large enterprises. However, regular web hosts like Mittwald, Webgo, Host Europe, etc., also offer domain configuration to some extent. Fig.1: Record entries for a domain of internetwarriors GmbH The Configuration of the Domain Name System The Domain Name System, often abbreviated as DNS, is one of the most important services on the internet. So-called DNS servers ensure that URL names are converted into IP addresses. When you type a URL into the browser, servers on the internet cannot initially do anything with that because they use IP addresses to identify themselves. An IP address is a unique numerical combination assigned individually to each device on the internet. This is most easily compared to a phone number assigned to every phone line. To know which server has requested a webpage (the URL typed into the browser), the browser first sends a request to a DNS server. This server maintains a large database that stores the IP address of the corresponding server for each domain, similar to a phone book where each name is paired with a phone number. In response to the request, the DNS server sends back the IP address of the corresponding server for the webpage to the browser. The browser can then directly place the request for the webpage with the actual web server. Types of Record Entries For the DNS server to know which IP address is behind a URL, this must be set in the domain configuration. There are various so-called record entries for different services. The most important ones are: NS A AAAA MX Name Server Record - NS The Name Server Record, often referred to as NS, is responsible for name resolution. This means resolving the names of services, e.g., domains, into computer-readable addresses. These addresses are the so-called IP addresses. Each of the entries has a so-called TTL (Time to Live). This determines how long an entry remains valid in the cache before it must be renewed. Typically, this value is 86400 seconds, which means 24 hours. The relocation of a domain to a new web server correspondingly takes 24 hours, as all global DNS servers must first be updated before the correct IP address is delivered to browsers. Address Record - A The Address Record or A Resource Record ensures that an IPv4 address is assigned to an entry on a DNS server. IPv4 addresses are IP addresses based on a four-octet system. This is greatly limited in the number of possible IP addresses and can no longer cover the required IP addresses for all devices connected to the internet, e.g., computers, smartphones, servers, etc. Nevertheless, it is still very common today. Address Record - AAAA The Address Record AAAA essentially provides the same functionality as the Address Record A. However, it is based on the so-called IPv6 addressing system, which is the successor of IPv4. It offers significantly more IP addresses and can therefore cover a much larger number of IP addresses and DNS entries. Mail Exchange Record - MX An MX Resource Record describes under which domain the corresponding email server can be reached. Through this, email programs can send and receive their emails. It is important to note that this entry must always include a fully spelled-out URL. What Can We Do for You? With our many years of experience in web hosting and domain management, we are happy to assist you in managing your website. Please contact us if you do not want to handle the technical management of your website or are planning a relaunch. We would be very pleased to discuss your individual support needs with you.
A/B Testing with Google Optimize
May 11, 2017

Thorsten
Abrahamczik
Category:
Web analytics

With Google Optimize , the search engine provider has introduced a new tool for conducting experiments on websites. Originally introduced as part of the Google 360 Suite, Google now offers the program, with a few restrictions, as a free version. This makes it easy for all marketers to use the tool for their own experiments. Fig. 1: The homepage of Google Optimize In addition to A/B tests, the program also supports multivariate and redirect tests. The following distinguishes the different types of tests: A/B Testing: This involves testing individual variations of the same webpage. Typically, the variations differ only in small parts, such as a different button color or a new call to action. Multivariate Tests: These tests work similarly to A/B tests, but in this case, several elements of a page are tested to find the best possible combination of elements. At the same time, it allows better investigation of user interaction between individual variations for conversion optimization. This quickly leads to a significantly larger number of variations. Redirect Tests: In these tests, separate pages with their own URLs are tested against each other. This way, different versions of entire pages can be effectively tested. Creating an Account and Container in Google Optimize To get started, users must open the homepage URL of Google Optimize . Upon initial opening, email subscriptions for (Tips & Recommendations, Product Announcements, and Market Research) should be considered, but they can also be declined. In the next step, the user must configure their account settings once. Fig. 2: One-time configuration of the Google Optimize account After making the changes visible in Figure 2, you can immediately begin setting up a website test. For the user, an account and a container are immediately created in which tests can be managed and configured. The setup is identical to the Google Tag Manager, which also relies on an account with individual containers. This significantly facilitates operation. Start with a First Test Before marketers begin creating a test, they must ensure that the website to be tested has many visitors. Only then can valid data be collected. If a website receives only a few visits per month, the evaluation of a test takes much longer to obtain statistically valid tests. In this case, only a few variations should be tested. In addition, marketers should conduct only small tests initially, such as changing a button color or swapping out text. This allows them to learn how to use the tool and understand how to build meaningful and effective tests. More complex tests can be created later. To start a test, the user must create a new experiment. Google offers templates for this, in which the user must select a name for the experiment, a URL for the page to be examined, and the type of test. Figure 3 shows the corresponding screen from Google Optimize. Fig. 3: These details can be used to create an experiment in Google Optimize. Work with Variations Once an experiment is created, the user can create so-called variations. These are slight alterations within the website. Regardless of the number of variations, each variation is always tested against the original version of the website. The marketer can also specify at this point how much traffic should participate in the test and how much traffic each variation should receive. By default, 100% of the traffic participates in an A/B test, and this traffic is evenly distributed across all variations. So, if there is the original version of the website and one variant version, each will receive 50% of the traffic. Fig. 4: For each test, a goal and a hypothesis must be set. After the variations are created, goals and descriptions must be set for the test. Examples of test goals include: Reducing bounce rates Increasing the number of page views Increasing the number of transactions Subsequently, the individual variations must be configured. A visual editor is used for this, allowing marketers to make directly visible changes to the website. For small changes, no knowledge of HTML, CSS, or JavaScript is necessary. For more complex changes involving HTML, CSS, or JavaScript, a general technical understanding of HTML and CSS is certainly required. For JavaScript changes, programmers should be consulted. Fig. 5: Google recommends installing the Google Chrome browser plugin for operating the visual editor of Google Optimize. To work with the visual editor, a browser plugin must first be installed. Google Chrome checks for the plugin and, as seen in Figure 5, suggests installation if necessary. Once the plugin is installed, the user can open the website and make adjustments. Figures 6 - 10 show how users can make adjustments: Fig. 6: By hovering the mouse, users select a webpage element. The individual elements are directly marked and highlighted by Google Optimize. Fig. 7: Changes to the selected element can then be made using the visual editor at the bottom right of the screen. Fig. 8: By clicking on "Edit Element" in the visual editor, further options can be selected, in this case, "Edit text". Fig. 9: Subsequently, the text of the H1 headline can be easily modified. Fig. 10: The bar at the top of the screen shows, among other things, which element the user is in (H1), how many changes have been made, and how each change appears on different device types such as desktop, tablet, and smartphone. Once the desired changes are made and saved, the appearance and behavior of the changes must be checked on each device. Quick errors may occur due to individual programming that can be avoided through extended checks. Linking Google Optimize with Google Analytics In the next step, Google Optimize must be linked with Google Analytics . For this purpose, the user selects a data view within the desired Google Analytics property. The user behavior data of this data view is then used to evaluate the test. This way, changes in bounce rates, the number of transactions, etc., can be considered in the experiment. Integration of Google Optimize via the Google Tag Manager In its developer area for Google Optimize, Google recommends using a modified Google Analytics code. This loads faster and prevents screen flickering caused by the dynamically made changes to the website. As a result, the user does not see upon page load that they are being shown a different variant. However, Google Optimize can also be integrated via the Google Tag Manager. In this case, the Google Tag Manager code should be placed as high up in the source code as possible. This is necessary to avoid possible screen flickering. The execution order of the different codes upon a page load is as follows: The user loads the page The Google Tag Manager code is executed The Google Optimize code is executed The Google Analytics code is executed Due to the use of the Google Tag Manager, there is a delay in execution, as Optimize can only be executed once the Tag Manager is loaded and executed. This is not the case when using a modified Google Analytics code, and the Google Optimize code can be executed immediately. As a result, on pages with many images or resources to load, the mentioned screen flickering can be reduced or completely avoided. In any case, Google Analytics should only be executed after the Google Optimize code, regardless of the use of the Google Tag Manager. A corresponding configuration must be set in the advanced settings of all Google Analytics tags on the website. Figure 11 shows the configuration of a Google Optimize tag in the Google Tag Manager. In this tag, essentially only the property ID of Google Analytics (also known as the UA number) and the container ID of the Google Optimize experiment must be entered. To comply with data protection, the IP address should also be anonymized. The trigger should be configured as precisely as possible to trigger the tag only when the corresponding page is called or the corresponding event is triggered. Fig. 11: Configuration of a Google Optimize tag in the Google Tag Manager. Defining the Target Audience and Timeframe Once the Google Optimize tag is published in the Google Tag Manager or the Google Optimize snippet is embedded in the page's source code, the further configuration of the experiment can proceed. For this, the target audience and duration of the test must be defined. In the free version, only the number of users participating in the test can be selected for the target audience. Currently, granular settings related to the target audience, such as age, gender, source of access, etc., are not possible. The timeframe can be individually set. Alternatively, the test can be started immediately and, if desired, ended immediately. Figure 12 shows the currently possible settings. Fig. 12: : Target Audience settings in Google Optimize. After an experiment, marketers can evaluate the results in the reporting area. Here, the data is displayed both in total and for each variation separately. In addition to standard data such as page views, the winner of the test is also displayed, including the improvements collected over the baseline. The individual variations are also checked against the set goals. This allows, in conjunction with further Google Analytics data, easy identification of which day and time each variation improved. Fig. 13: Evaluation of the results of an experiment. Source: https://support.google.com/360suite/optimize/answer/6218117?hl=en. What We Can Do for You Would you like to test individual elements of your website to increase conversions, leads, or transactions? Start with the free version of Google Optimize to quickly and easily conduct individual experiments. We are happy to advise you on the implementation and evaluation of corresponding tests for conversion optimization, in connection with your set website goals. Contact us.
What clients can expect from your agency in web analytics support
Feb 2, 2017

Thorsten
Abrahamczik
Category:
Web analytics

Web analytics, what exactly is it? For many of our clients, web analytics means using a tool that receives data that is barely interpretable and possibly invalid. The operation quickly becomes overwhelming, so the tool is used only once every few months. This happens whenever management wants to see data about the website or when it's necessary to justify why the planned marketing budget for the next year is so high. But web analytics is much more than that! Web analytics means questioning interactions on your own website, evaluating the behavior of individual target groups, creating customer journeys, and much more. Ultimately, from all these evaluations and results, an action plan should be created that enables you to specifically optimize your website. In this article, we would like to explain what you, as an online marketer, can expect from an agency that supports you in the area of web analytics. We will discuss our experiences and explain how we approach the topic of web analytics with a client. Fig. 1: Cross-device web analytics | Source: http://bit.ly/2kxlWF4 Phase 1 – Reviewing the Status Quo in Tracking If you approach us as a potential new customer to talk about web analytics, we first describe what web analytics means to us: Web analytics is the foundation of all online marketing activities and provides the basis for all marketing-related activities, both online and offline. It collects both quantitative and qualitative data. When used properly, web analytics describes what happens due to customer interactions and, most importantly, why these interactions occur. Furthermore, it places your own data, if available in benchmarks, in relation to the data of other companies or industries. This shows where you can improve as a company. But more than anything, web analytics is one thing: continuous. Web analytics is very comprehensive and requires intensive support. However, it is not rocket science and can be easily implemented. To ensure this, we first check your current tracking implementation. This way, we obtain a status quo of your web analytics and can better assess where problems and potentials exist. Reviewing the Configuration of the Web Analytics Tool First, we go into the tracking tool and check which data is flowing in, how the tool is configured, and whether the data is valid. Most of our clients use Google Analytics as their tool, making the verification relatively straightforward and almost standardized. Nevertheless, we also look for peculiarities in the data and configuration. Examples include: Are filters being used correctly? Is there spam in the data that distorts evaluations? Is Google Analytics linked with other services like AdWords, Search Console, etc.? Are the data from internal searches being collected? Are demographic data activated? Have goals been set up in Google Analytics? As you can see, there is much to consider when configuring Google Analytics. Especially if you feel uncertain as a marketer and cannot precisely assess which setting causes which impact, key figures can easily be collected or interpreted incorrectly. A classic example is excluding your own accesses through an IP address filter. When we check the filter configurations in Google Analytics, it is incorrectly set up and does not function in 95% of the cases. Our staff member Bettina Wille has written an extensive article about what you need to watch out for when configuring filters in Google Analytics. We also check if you are using reports. These are generally represented in dashboards, radar events, or custom reports, so verification is easily feasible. Checking the Technical Implementation in the Source Code Once we know how your Google Analytics is configured, we check the implementation in the source code. This is not only about seeing whether the tracking code is implemented but also about how it is implemented. Below is a sample selection of aspects we check: Which version of the Google Analytics tracking have you implemented? Has the tracking code been fully implemented according to the configured Google Analytics settings? Are there pages where the tracking code is not implemented? Are multiple tracking code implementations present so that data is collected twice? Are specialties such as cross-domain tracking, e-commerce tracking, etc., implemented correctly? Are additional tracking features being used, e.g., custom dimensions, event tracking, etc.? Are referrers being correctly passed on and is direct traffic indeed just direct traffic? Reviewing Data Protection Finally, we check in individual areas whether your Google Analytics is installed in compliance with data protection regulations. Please note, however, that we are not a law firm and, therefore, cannot provide legally binding advice. Nonetheless, there are features that can be easily checked, such as: Are you anonymizing the user's IP address? Do you have a privacy page that mentions the use of Google Analytics? Is there an option for users to opt out of tracking, both via a browser plugin and a functioning opt-out cookie? Have you signed a data processing agreement with Google? Phase 2 – Developing a Tracking Concept In this article, we have so far focused exclusively on implementing with Google Analytics. Of course, there are also other types of web analytics with additional tools for things like A/B testing, surveys, etc., which also need to be reviewed. However, we do not wish to go into these tools individually in this article. After we have worked out a precise overview of your current tracking and activities in the area of web analytics, we move on to developing a tracking concept. In this, we describe your current status quo and define in great detail what tracking should be used in the future. We also determine jointly with you which data should be collected beyond standard tracking, which website goals should be examined, which KPIs should be defined, and how the configuration should be implemented. Figure 2 shows an excerpt from the table of contents of a tracking concept: Fig. 2: Together with you, we develop a tracking concept. Analysis and Research of Tracking Opportunities by the Agency At the beginning, we explore, based on the status quo analysis, possibilities for enhanced tracking on your website. It is important for us to only collect data that is valuable to you. We can also recommend tracking where we track everything and nothing, so to speak. In this case, we collect a lot of data, which leads to the point where you can no longer analyze it because the data volume is simply too large. Of course, that is not our goal. From our perspective, it requires a precise measure where you can work well with the key figures and receive all the information you need for your evaluations. Aligning Tracking Goals with the Client For the reasons mentioned above, we engage in very intensive exchanges with you during this phase. Here, there are many conversations with different people in your company to understand your requirements for web analytics better. Some of the topics we discuss with you include: Company goals Website objectives KPIs already in use Previous internal reports Hierarchy levels in the company including different reports for different contacts Cooperations with third-party providers Differences between reporting and web analytics Motivation for using Google Analytics Defining KPIs and Goals for the Web Analytics Through these discussions, we not only learn about your requirements but can also better assess which topics in web analytics are particularly interesting for you. Based on this, we can provide targeted recommendations for tracking implementation, KPIs, goals, etc. At this point, you do not yet know our final draft/proposal. Discussion of the Draft Concept with the Client and Approval from the Client Once we have worked out your tracking concept, there is a joint meeting with you in which we discuss the tracking concept with you in detail. This is especially important as the concept is partly very technical. However, it is essential that you completely understand the concept. If you have any requests for changes at this stage, we will discuss them and, if necessary, incorporate them into the concept. Phase 3 – Implementation of the New Tracking Once you have approved the tracking concept, we begin implementing the tracking. Here, we always start with the technical implementation before starting the configuration of Google Analytics. Using the Latest Technologies When we receive an order to implement tracking, we naturally always use the latest technologies. This includes using the Google Tag Manager as well as Universal Analytics. If specific reports are desired, we recommend the client use Google Data Studio. Google Tag Manager As we have already written in previous articles about the Google Tag Manager, the entire tag management can be easily handled, both for Google Analytics and for other tools like AdWords, third-party providers, etc. A major advantage of the Tag Manager is that in very few cases does the IT need to adapt the source code. The tasks associated with managing tags can then be carried out directly by the marketer in the tool. We have already described the functionality of the Google Tag Manager in a comprehensive article. Universal Analytics Universal Analytics is the current version of Google Analytics. The data collection of Universal Analytics differs slightly from the old asynchronous Google Analytics and offers additional benefits such as Enhanced E-Commerce, UserID, etc. Google announced in 2015 that they would cease to support asynchronous tracking in the future. When support will be discontinued is still unclear. Data Studio and Other Offerings With the Google Data Studio, another tool from the Google 360° Suite has been made available as a free version. Since the beginning of February, the tool allows you to create as many reports per email address as you like, the only limitation being that the Double-Click connector cannot be used in the free version. Otherwise, it has the same functionality as the paid version of the 360° Suite. Particularly interesting for marketers is that the reports can be provided with their own CI. This is especially useful when the reports are forwarded to management. Collaboration with IT or the IT Service Provider Once we start implementation, usually only a few adjustments need to be made to the source code. The old tracking remains intact for the time being. This ensures that the main tracking is not affected. For all changes to the source code, the IT or service provider receives precise instructions from us on what needs to be changed. Generally, the Google Tag Manager code needs to be embedded, but often also the opt-out cookie. If e-commerce is being used or conversion values should be dynamically transmitted from the website, adjustments are also required for this. The entire configuration of your new tracking is then implemented by us in the Google Tag Manager. This means we set up the Universal Analytics tag and create variables, triggers, and other tags to fulfill the specifications from the tracking concept. If we also support the client in other online marketing areas like Google AdWords, we implement this directly in the Google Tag Manager as well. This allows for easier management in the future. Throughout the entire phase, we work closely with you and your IT to ensure a correct implementation. If questions arise on the IT side, we offer advisory support. If we maintain your website/content management system, the adjustments are naturally carried out by us. Configuration of the Used Tools With a delay to the basic setup of Universal Analytics in the Google Tag Manager, we set up a so-called test property in Google Analytics that we use to test the data we collect. If we did not do this, we would have to direct the new data into your main property. This would lead to data distortion and problems with clarity that we want to avoid. Only when we know that all data is correctly transmitted from the Google Tag Manager to Google Analytics do we configure the Google Analytics of the main property. This is because some configurations are only possible when certain tags are configured in the Google Tag Manager. Once the data is correctly transmitted by the Google Tag Manager, and Google Analytics is correctly configured, we inform IT that they can remove the old tracking code. Once this has occurred, we adjust the tracking in the Google Tag Manager so that the data no longer flows into the test property but into your main property. This way, we ensure that your old metrics in Google Analytics are not lost and that you can compare your new data with the old data. It should be noted that your old data may not be valid. Setting Up Reporting Part of the configuration of Google Analytics and Google Data Studio may include setting up reports. These are created according to the specifications of the tracking concept. Client Training in the Use of the Tools Once the entire tracking has been implemented, you need to understand and be able to apply the entire implementation independently. For this reason, we conduct an introduction to the implementation, depending on your previous knowledge. This includes all tools (Google Tag Manager, Google Analytics, and Google Data Studio). This is purely an introduction to the implementation. If you as a client have very little experience in Google Analytics and, for example, no experience with Google Tag Manager or Google Data Studio, we recommend a thorough web analytics training. In this, we explain to you not only all the tools but also the basics of web analytics. Depending on the scope, this lasts between one and two days. Learn more about our Google Analytics training sessions. Phase 4 – Ongoing Support So that the new metrics from Google Analytics don't go unused after setup, we support our clients after implementation with ongoing support. This is generally done in two ways. On the one hand, we ensure that the data will continue to be collected correctly. On the other hand, we conduct more in-depth analyses. Regular Analyses and Creation of Evaluations by the Agency This allows you to focus on your reports while we conduct comprehensive web analytics for you. Here, we focus specifically on individual subareas. For example, we analyze the behavior of individual target groups, examine the behavior of users in specific areas of the website, or, alongside a custom channel attribution, develop a basis for an attribution model. In a second step, we further develop this together with you so that you can understand exactly how users come into contact with your company and how you might optimize individual channels to target users more precisely. We also support you in areas such as A/B and multivariate testing. As these tasks are time-consuming but also very productive, our clients gladly take advantage of our expertise and leave these tasks to us. Continuous Support in Case of Problems This topic primarily involves technical support. It often happens that errors in data collection occur due to website adjustments. This happens, for example, through renaming individual buttons, removing website elements, or a complete adjustment of a website element. Generally, IT doesn’t consider during implementation that changes might affect tracking. These errors usually become apparent to us quickly, and we can correct them in consultation with you. What Can We Do for You? Would you like to check your tracking, significantly expand it, or simply outsource the evaluation of the data to a third party? Do you want to conduct targeted optimizations with your web analytics and thus increase your leads? Contact us , and we will be happy to advise you on how we can optimally collaborate with you.
Check Google Analytics Implementation with Screaming Frog
Sep 15, 2016

Thorsten
Abrahamczik
Category:
Web analytics

Modern websites and content management systems have become very complex, making it challenging for users to make technical adjustments. From our day-to-day business, we know that online marketers often encounter issues, especially with tracking. But only through valid tracking can you generate clean data in the analytics tools. Only then are qualitative analyses possible. In this article, we present a method that allows you to easily check whether the Google Analytics code is installed on all pages. Furthermore, we will explain how to verify the code implementation. Screaming Frog – The Tool for SEO A well-known SEO tool is the Screaming Frog program. The main task of the tool is to crawl websites. For this purpose, a crawler (also referred to as a bot) is sent to the respective website to gather information on all subpages and their contents. Marketers can thus easily check to what extent the website has potential for SEO. In the free version, up to 500 website elements (HTML pages, images, CSS files, etc.) can be analyzed. The paid version is much more powerful and crawls all pages. Besides standard reporting, the tool offers the possibility to connect with external services such as Google Search Console and Google Analytics to obtain even more accurate analyses. Additionally, it also provides the option to conduct investigations on the website with custom filters. Essentially, this involves two different methods: Search: The affected pages are listed with URLs in Screaming Frog. Extraction: In this case, the desired content of the affected pages is displayed by Screaming Frog. Configuration of Screaming Frog Figure 1 shows Screaming Frog immediately after opening the program. By typing a URL in the bar and clicking on Start, the corresponding page or domain is immediately crawled, and initial information about the URLs flows into almost all tabs. The "Custom" tab will be important for you later. Fig. 1: The structure of Screaming Frog Identifying Pages Where the Google Analytics Code Is Missing If you want to identify pages or URLs where, for example, the Google Analytics code is not embedded, click on the "Search" button as shown in Figure 2. Fig. 2: Using custom filters in Screaming Frog In the subsequent dialog, Screaming Frog offers you ten different filters. To search for pages where Google Analytics is not installed, you only need one filter. Set the first field to "Does Not Contain" and enter the UA number of your Google Analytics property. You can also make other inputs there, but it's important that it’s something unique from the Google Analytics code. Also ensure that the entry does not match any other element in the source code. From our point of view, the Google Analytics UA number is a good choice. Fig. 3: Using the search filter in Screaming Frog If you want to check whether the Google Tag Manager is correctly embedded, you could enter "GTM-XXXXXX" as the container's ID. In this case, you are also using an element unique to the Google Tag Manager. Subsequently, under the "Custom" tab, all pages where the corresponding search term is not found will be listed. This way, you can easily identify which pages still need adjustments to achieve complete tracking. You can also go to IT with a specific action plan in this manner. Checking the Google Analytics Property ID on All Pages In the event that Google Analytics is installed on all web pages, you should additionally verify that the correct UA number is used. Mistakes can easily occur, and then you also don't have valid tracking. The "Extraction" method is perfect for this. Click on "Extraction" as shown in Figure 2. In the following dialog (Figure 4), enter "Google Analytics UA number" as an example in the first field. This field is used to name the corresponding column in the "Custom" tab. Fig. 4: Setting an extraction filter to read the value of the Google Analytics UA number In the second field, select "Regex". Regex is short for "regular expressions" and provides the opportunity to identify exact letter, number, and character combinations within a given text area, e.g., the source code of a webpage. To find Google Analytics elements on the page, use the following regex: ["'](UA-.*?)["'] This way, you can see which Google Analytics UA number is used for each page of your domain. You can easily spot any typos and correct them afterward. For the Google Tag Manager, you would use "["'](GTM-.*?)["']" as a regex filter in this case. In the next figure, you can see how the results are then displayed in the "Custom" tab. Fig. 5: "Custom" tab Conclusion Screaming Frog offers strong expandability with custom filters besides its diverse SEO analysis capabilities. Here, users can also check non-SEO content and gain insights that greatly simplify daily work. What Can We Do for You? Are you unsure if Google Analytics is correctly embedded on your site or wonder if Google Analytics tracking can be extended beyond standard tracking? Do you want to measure conversions even more accurately? Contact us, and we will be happy to advise you on checking existing tracking, creating a tracking concept, or implementing tracking. We look forward to your inquiry .
Set filters in Google Analytics
Jul 28, 2016

Thorsten
Abrahamczik
Category:
Web analytics

Google Analytics is the most popular web analysis tool. As a website operator, you can understand the behavior of your users on the site and subsequently define appropriate measures for page optimization. To ensure that you only find relevant data in your evaluation, it's wise to set certain filters. Why are filters useful in Google Analytics? Google Analytics captures all user data on the website without filtering. However, it is not always 100% effective to use all values for an analysis. With the help of filters, you can initially channel the data through a kind of sieve so that your data view only shows the values you need. Generally, filters can be categorized into the following main categories: Inclusive filter: only a defined filter pattern gets through Exclusion filter: the defined filter pattern is excluded Search & Replace: designations recorded by Google (e.g., Homepage = "/") are rewritten Lowercase: Reduction of duplicate results by using all lowercase letters A common example of setting a filter is the exclusion of an IP address range. This allows you to exclude data from your company or agency network if it uses a static IP. The same applies to your IT service provider who regularly visits the site with a static IP address, without it being relevant for site optimization. This filter is not possible with dynamic IP addresses as they change every 24 hours. Access from spam or bots can partly be excluded by filters. Many referral URLs are now known and published on portals that go to the site as spam and skew your traffic. Check the referrals that lead users to your website. It is typical to see 1 page/session, 0:00 time on site, and an almost 100% bounce rate. When researching, you can get an overview of the referral URLs via the left navigation bar under Acquisition, All Traffic, Referrals. Fig. 1: Display referrals in Google Analytics But you can also detect spam behavior in the reports of Google Analytics under "Locations", "Website Content", and "All Traffic". A common example of spam access is the following: Fig. 2: Example of spam access How to exclude the IP address range for your corporate network If you want to check user behavior with the web analysis tool Google Analytics, accesses from your own network are obstructive. Internal page hits should not be equated with the character of a browsing user. To prevent accesses from your own ranks from appearing in your data, set a filter as follows: Navigate to your Google Analytics account and open the admin view: Fig. 3: Admin view in Google Analytics In the right column of the data view, you will already see the "Filter" field. Here you can now add a new filter. Configure the filter. Give it a meaningful name, select "predefined" as the filter type, and use an "Exclude" filter. Choose the "begins with" command and enter the start of the IP address to be excluded in the last field. You should never use the "equals" command. Due to the data protection-compliant anonymization of the IP address, the fourth/last part (the last 3 digits) of the IP address is always truncated, so the complete sequence of numbers never reaches Analytics. Fig. 4: Filter to exclude IP addresses How to exclude referral URLs If you found out through checking your referral URLs that certain accesses are done by bots and thus flow into your analysis as spam, you should create a filter to exclude this data. First, collect a list of these URLs or research a list and also create a custom "Exclude" filter. Give it a meaningful designation and use the referral as the filter field. Then enter all the researched URLs into the filter pattern. Separate the individual URLs with a pipe (|). Add a backslash () before EACH dot, so the dot does not execute a command as a regular expression. Avoid spaces in the filter pattern entirely. Unfortunately, Google Analytics provides only 255 characters in the filter pattern, so you may need to set up multiple filters to exclude referrals. Test filters before applying Once you've set the filters, they are immediately active. However, you cannot regain filtered data or reverse the sieving effect. We therefore recommend testing the filters in a test view before applying them to the live environment. Also use a third data view that collects all your website data without filters. This way, you can still reconstruct the correct values in case of possible errors. How you should configure the data views can be found in our article Google Analytics Basic Configuration – What You Should Pay Attention To . What we can do for you The proper use of filters cleans up your data so that you can draw meaningful results from analysis with Google. If you need support with Web Analysis , we look forward to your inquiry.
The recording feature of Google Tag Assistant
Jun 30, 2016

Thorsten
Abrahamczik
Category:
Web analytics

Many companies incorporate Google Analytics, a free web analysis tool, into their website to get an overview of their website's visitor numbers. However, this is often where the problems begin. In many cases, the tool or its code is simply copied into the page's source code. This does not take into consideration that the Google Analytics code must be placed in a specific spot, nor that the IP address needs to be anonymized. It is also seldom considered that in certain cases the referrer, meaning the reference/link through which a user came to your site, is not correctly passed on. In this case, the website visitor is recorded as a direct entry and not as a visit via a referral. Therefore, users are working with skewed numbers. To identify and avoid these issues, Google offers a practical browser extension for Chrome called Google Tag Assistant, as mentioned in the blog article The Best Tools for Successful Entry into Web Analysis . The Functions of Google Tag Assistant First, Google Tag Assistant provides the ability to display information about all Google tags. These can typically include the following tags, as shown in Screenshot 1, which also shows the button for the record function. Google Analytics, Remarketing, Conversions Tracking, DoubleClick, etc. Figure 1: Overview of Google Tag Assistant For each individual tag, you can display detailed information about the current status of the tag. A color-coding system further indicates whether the tag is correctly implemented (green), if there are slight deviations from the implementation recommendations (blue or yellow), or if there are significant problems causing errors in tag execution (red). The level of detail in the provided information can be configured in the Tag Assistant settings for each tag type. The Record Function As illustrated in screenshots one and two, a record function has been added to the tool with Google Analytics in mind, which allows for precise analyses across multiple page views. This enables you as a user to easily determine whether your data is being correctly captured and processed for Google Analytics. The advantage of this method: data is measured individually for all executed Google Analytics tags. In addition to page views, you can, for example, also examine event tracking or e-commerce tracking. This way, you can find errors that you would otherwise only uncover with significantly greater effort. Figure 2: Recording has begun Using this, we were able to discover an error in a customer's cross-domain tracking. It was correctly configured for page view tracking, but not for several event tags. This led to the referrer being correctly passed during page views of the second domain, but not for any event tags, where the fallback "Direct/None" was used. This verifiably distorted the metrics in Google Analytics. After a corresponding adjustment to the event tags, the appropriate referrers were correctly passed on, and the metrics were correctly integrated into Google Analytics once more. An advantageous feature is the ability to directly link Tag Assistant recordings with Google Analytics. In the analysis area of Google Tag Assistant, you can selectively choose individual data views of the Google Analytics account. You can also send specific location data to check whether certain data, for instance due to IP address filters, does not integrate into Google Analytics. In such cases, you can make the necessary adjustments in your Google Analytics configuration and immediately update the recording with a click of the refresh button. A new recording is not needed to see the results of the adjustments immediately. Screenshot 3 shows a section of the analysis area: Figure 3: Overview of the Google Analytics Record Function Analysis in Google Tag Assistant What We Can Do for You At internetwarriors, the Google Tag Assistant with its record function is one of the standard tools in the field of web analysis and SEA. We would be happy to review your Google implementation as well and identify any potential errors. If desired, we can also fix them. Improve the quality of your analyses with more accurate metrics and, correspondingly, your marketing budget allocation. Contact us.
Show more posts
Was Onlineshops vom E-Rezept-Boom der Versandapotheken lernen können
Sep 29, 2026

Axel
Zawierucha
Category:
Search Engine Optimization

Kurz zusammengefasst: Medikamente sind die am schnellsten wachsende Warengruppe im deutschen Onlinehandel: plus 13,9 Prozent auf 547 Millionen Euro im zweiten Quartal 2026. Stationäre Apotheken wuchsen im selben Zeitraum nominal um 3,2 Prozent. Der Treiber ist das E-Rezept, das seit Januar 2024 Pflicht ist. Versandapotheken gewinnen mit vier Dingen, die jeder Onlineshop übernehmen kann: sichtbares Vertrauen, einfacher Checkout trotz Regulierung, automatisierte Nachkäufe und die Verbindung von online und lokal. Wenn mich jemand fragt, welche Branche im deutschen Onlinehandel gerade am meisten richtig macht, sage ich: die Versandapotheken. Das überrascht viele, weil kaum ein Markt so streng reguliert ist. Aber genau das macht ihn so lehrreich. Wer es schafft, verschreibungspflichtige Medikamente per Smartphone zu verkaufen, hat Probleme gelöst, die Händler in anderen Branchen oft gar nicht erst angehen. Warum wachsen Versandapotheken so stark? Weil das E-Rezept die größte Hürde beseitigt hat, die es beim Onlinekauf von Medikamenten gab: das Papierrezept, das man per Post einschicken musste. Seit dem 1. Januar 2024 ist das E-Rezept für gesetzlich Versicherte Pflicht, am 17. Oktober 2025 wurde das milliardste eingelöst. Der Unterschied zwischen den Kanälen ist beachtlich. Online-Apotheken setzten im zweiten Quartal 2026 mit Medikamenten 547 Millionen Euro um, 13,9 Prozent mehr als im Vorjahr, laut bevh das stärkste Wachstum seit Einführung des E-Rezepts. Stationäre Apotheken wuchsen laut Destatis im selben Zeitraum nominal um 3,2 Prozent und real um 1,6 Prozent. Online wächst also mehr als viermal so schnell. Welche Warengruppen wachsen online am stärksten? Warengruppe Online-Umsatz Q2 2026 Wachstum zum Vorjahr Medikamente 547 Mio. € +13,9 % Drogerie 974 Mio. € +11,7 % DIY und Blumen 1.176 Mio. € +10,9 % Tierbedarf 641 Mio. € +9,8 % Lebensmittel 1.155 Mio. € +8,9 % Möbel, Lampen und Dekoration 1.366 Mio. € minus 2,1 % Quelle: bevh, Q2 2026 Das Muster ist klar. Was man regelmäßig braucht, wandert ins Netz. Bei großen Anschaffungen wie Möbeln halten sich die Leute zurück. Wer Versorgungsgüter verkauft, hat also gerade Rückenwind, muss aber auch damit rechnen, dass Kunden sehr genau vergleichen. Wie bauen Versandapotheken Vertrauen auf? Sie machen Vertrauen überprüfbar. Medikamente gehören für Google zur sensibelsten Kategorie, den sogenannten YMYL-Themen („Your Money or Your Life“). Dort bewertet Google Inhalte besonders streng nach Erfahrung, Expertise, Autorität und Vertrauenswürdigkeit, abgekürzt E-E-A-T. Dieselben Signale nutzen KI-Systeme, wenn sie entscheiden, welche Gesundheitsquelle sie zitieren. Versandapotheken müssen im Versandhandelsregister des BfArM eingetragen sein und auf jeder Seite, auf der sie Arzneimittel anbieten, das EU-Sicherheitslogo zeigen. Ein Klick darauf führt direkt zum Registereintrag. Dazu kommen Ratgeber, die von approbierten Apothekerinnen und Apothekern gezeichnet sind, und Hinweise auf mögliche Wechselwirkungen direkt im Bestellprozess. Was andere Shops daraus lernen können: Zeigen Sie Prüfsiegel, Zertifikate, Herkunft und Expertennamen dort, wo die Kaufentscheidung fällt, nicht im Footer. Jede Aussage, die der Kunde selbst nachprüfen kann, senkt die Abbruchquote. Und sie erhöht die Chance, dass ChatGPT oder Google AI Mode Sie als verlässliche Quelle nennen. Wie wird ein komplizierter Prozess einfach? Die Apotheken haben den regulierten Ablauf in mehrere digitale Wege übersetzt. Online lässt sich ein E-Rezept über die App der gematik einlösen, über das CardLink-Verfahren in der App der Versandapotheke oder per Ausdruck mit Rezeptcode. Beim CardLink-Verfahren hält der Kunde seine Gesundheitskarte ans Smartphone, der Rest passiert im Hintergrund. Im Lebensmittelhandel passiert gerade etwas Ähnliches beim Bezahlen. Rewe hat Bonus und Bezahlen seit Juli 2025 in einer App zusammengelegt, Edeka bietet Payback Pay, Lidl eine Bezahlfunktion in Lidl Plus. Bar bezahlt wird im LEH nur noch rund ein Drittel des Umsatzes, bei Discountern ist der Baranteil seit 2019 von 56 auf 35 Prozent gefallen. Das Learning für jeden Shop ist simpel, wird aber erstaunlich selten umgesetzt: Jedes überflüssige Formularfeld kostet Umsatz. Wenn Apple Pay, Google Pay oder PayPal die Adresse ohnehin mitliefern, muss der Kunde sie nicht nochmal eintippen. Und wer Zahlung, Kundenkonto und Treueprogramm in einem Schritt bündelt, macht den nächsten Kauf leichter als den ersten. Wie wird aus einem Einmalkauf ein Nachkauf? Versandapotheken haben Routinekäufe systematisiert. Wer ein Dauermedikament braucht, bekommt rechtzeitig vor dem Aufbrauchen eine Erinnerung und bestellt mit einem Klick nach. Folgerezepte im selben Quartal gehen inzwischen ohne erneuten Praxisbesuch. Das lässt sich auf viele Produkte übertragen. Bei Kosmetik, Tierfutter, Kaffee, Filtern oder Druckerpatronen kann man den Nachkaufzeitpunkt ziemlich genau ausrechnen: Packungsgröße geteilt durch den üblichen Verbrauch. Eine automatische Erinnerung per E-Mail oder App zum richtigen Zeitpunkt ist einer der wirksamsten Hebel für den Kundenwert, den ich kenne. Und dass Bindung sich lohnt, zeigt auch der Hello-Again-Loyalty-Report 2026: Nutzer von Treueprogrammen geben im Schnitt 30 Prozent mehr aus, 80 Prozent erwarten Vorteile, die sie sofort einlösen können. Wie verbinden die Erfolgreichen online und lokal? Rezept digital schicken, Medikament per Bote bekommen oder in der Apotheke um die Ecke abholen, das funktioniert, weil online und lokal sich ergänzen. Der Lebensmittelhandel geht denselben Weg. Rewe bietet Click & Collect in mehr als 2.000 Märkten an und erreicht mit Liefer- und Abholservice nach eigenen Angaben 98 Prozent der Haushalte. Seit März 2026 testet Rewe mit „Drive & Go“ die Übergabe direkt ans Auto in unter drei Minuten. Für Händler mit Filialen heißt das: Zeigen Sie, was vor Ort verfügbar ist. Die Werkzeuge dafür sind ein gepflegtes Google-Unternehmensprofil, ein lokaler Inventarfeed im Google Merchant Center als Grundlage für Local Inventory Ads und strukturierte Daten für die Filialseiten. Wer akuten Bedarf hat, kauft beim, der sofort liefern kann. Worauf Sie bei Gesundheitsthemen achten müssen Ein Hinweis, der mir wichtig ist: Werbung für Arzneimittel und Gesundheitsaussagen unterliegen dem Heilmittelwerbegesetz. Was bei Kaffeebohnen ein lockerer Werbespruch ist, kann bei Nahrungsergänzungsmitteln oder Medizinprodukten eine Abmahnung auslösen. Lassen Sie solche Aussagen vor der Veröffentlichung rechtlich prüfen. Mein Fazit Das E-Rezept hat die Versandapotheken zur am schnellsten wachsenden Warengruppe im deutschen Onlinehandel gemacht, mit einem Wachstum, das den stationären Kanal um das Vierfache übertrifft. Der Erfolg hat wenig mit Pharma zu tun und viel mit gutem E-Commerce: Vertrauen, das man nachprüfen kann, ein Checkout ohne Hürden, Nachkäufe, die fast von allein passieren, und die Verbindung von online und lokal. Jeder Onlineshop, der diese vier Dinge ernst nimmt, verbessert Conversion, Kundenbindung und seine Sichtbarkeit in Suchmaschinen und KI-Antworten zugleich. Häufige Fragen Wie löse ich ein E-Rezept bei einer Online-Apotheke ein? Über die E-Rezept-App der gematik, über CardLink in der App der Versandapotheke oder per Ausdruck mit Rezeptcode. Das Einstecken der Gesundheitskarte funktioniert nur in der Apotheke vor Ort. Woran erkenne ich eine legale Online-Apotheke? Am EU-Sicherheitslogo. Ein Klick darauf muss zum Eintrag der Apotheke im Versandhandelsregister des BfArM führen. Wie schnell wachsen Online-Apotheken im Vergleich zu stationären Apotheken? Im zweiten Quartal 2026 wuchs der Online-Umsatz mit Medikamenten um 13,9 Prozent, stationäre Apotheken legten nominal um 3,2 Prozent zu. Was bedeutet E-E-A-T für Shops außerhalb der Gesundheitsbranche? Auch bei Produkten mit Bezug zu Sicherheit, Gesundheit oder Finanzen bewertet Google Expertise und Vertrauenswürdigkeit. Namentlich genannte Experten, Quellenangaben und Zertifikate stärken beides. Sie wollen die Conversion Ihres Shops steigern und Ihre Sichtbarkeit auf E-E-A-T-Niveau bringen? Wir begleiten Sie bei Conversion-Optimierung, technischem SEO und GEO. Shop-Audit anfragen Quellen: bevh, Pressemitteilung vom 06.07.2026; KPMG Retail Sales Monitor 3/2026 (auf Basis Destatis, EHI Retail Institute, Hello-Again-Loyalty-Report 2026); gematik, Meldung vom Oktober 2025; BfArM, EU-Sicherheitslogo; it-recht-kanzlei.de
Marktplatz oder eigener Shop? Was D2C-Marken 2026 wirklich wissen müssen
Sep 29, 2026

Axel
Zawierucha
Category:
Growth Marketing

Kurz zusammengefasst: Online-Marktplätze setzten im zweiten Quartal 2026 11,5 Milliarden Euro um, das sind 54 Prozent des deutschen E-Commerce. Herstellerversender wuchsen mit 6,3 Prozent fast genauso schnell, klassische Onlineshops nur um 3,8 Prozent. Für die meisten Marken ist die Antwort deshalb nicht „entweder oder“: Der Marktplatz gewinnt Neukunden, der eigene Shop macht sie profitabel. Die unterschätzte Konkurrenz sitzt dabei nicht nur bei Amazon und Temu, sondern im Regal des Händlers, als Handelsmarke. Die Frage, die ich von Markenherstellern gerade am häufigsten höre, lautet: Brauchen wir unseren eigenen Shop überhaupt noch? Meine Antwort ist ja. Aber nicht mehr für dieselbe Aufgabe wie vor fünf Jahren. Damals war der eigene Shop der Wachstumsmotor. Heute ist er der Ort, an dem aus Kunden Stammkunden werden. Das Wachstum kommt woanders her. Wie verteilt sich der deutsche Onlinehandel? Die Zahlen des bevh für das zweite Quartal 2026 sind ziemlich eindeutig: Versendertyp Umsatz Q2 2026 Wachstum zum Vorjahr Online-Marktplätze 11,52 Mrd. € +6,4 % Herstellerversender (D2C) 742 Mio. € +6,3 % Klassische Onlineshops 6,06 Mrd. € +3,8 % Multichannel-Händler mit stationärem Ursprung 2,74 Mrd. € +2,5 % E-Commerce gesamt 21,19 Mrd. € +5,1 % Quelle: bevh, Q2 2026 Die interessanteste Zeile ist für mich nicht die erste, sondern die dritte. Marktplätze wachsen, das wissen alle. Aber Hersteller, die direkt verkaufen, wachsen fast genauso schnell. Wer unter Druck gerät, sind die klassischen Onlineshops ohne eigene Marke und ohne Plattformgröße. Die sitzen genau zwischen den Stühlen. Dazu kommen die asiatischen Plattformen. Temu, Shein und AliExpress wachsen um mehr als 20 Prozent und machen inzwischen 5,3 Prozent des gesamten Onlineumsatzes aus, ein Jahr vorher waren es 4,9 Prozent. Im Online-Modehandel entfallen schon mehr als 16 Prozent aller Bestellungen auf sie. Vom neuen 3-Euro-Pauschalzoll auf Kleinsendungen seit 1. Juli 2026 erwartet der bevh wenig, weil die Anbieter längst eigene Lager in Europa aufbauen. Im Modehandel sieht man die Verschiebung besonders deutlich. Die stationären Modehändler im TW-Testclub verloren im ersten Halbjahr 2026 4,0 Prozent Umsatz, die Online-Modeausgaben stiegen im selben Zeitraum um 4,0 Prozent. KPMG ordnet das nicht als reinen Kanalwechsel ein, sondern als Marktanteilsgewinn der Marktplätze. Warum gewinnen die Marktplätze? Aus drei Gründen, die ein einzelner Shop kaum ausgleichen kann. Der erste ist Bequemlichkeit: Die Zahlungsdaten sind hinterlegt, die Rückgabe läuft nach einem bekannten Muster, und bei schwacher Konsumstimmung registriert sich niemand gern in einem neuen Shop. Der zweite ist Technik: Große Plattformen investieren Milliarden in Empfehlungen und Preissteuerung. Der dritte ist Logistik: Lieferung am nächsten Tag setzt einen Standard, der im Eigenversand teuer wird. Warum wachsen D2C-Marken trotzdem? Weil eine starke Marke einen Grund liefert, direkt beim Hersteller zu kaufen. Das kann ein exklusives Produkt sein, eine Personalisierung, ein Abo oder einfach das Gefühl, beim Original zu kaufen. Wer diesen Grund hat, wächst auch gegen Amazon. Wer ihn nicht hat, verliert. Warum ist die Handelsmarke die unterschätzte Konkurrenz? Das ist für mich der Punkt, der in fast jeder D2C-Diskussion fehlt. Handelsmarken sind längst keine Billigalternative mehr. Ihr Umsatzanteil im Lebensmittelhandel stieg laut YouGov von 40 Prozent im Jahr 2021 auf 46 Prozent im Jahr 2025, das vierte Jahr in Folge. Premium-Handelsmarken wuchsen im selben Zeitraum von 4 auf 6 Prozent Marktanteil. Bei verpackten Bioprodukten liegt der Eigenmarkenanteil schon bei 67 Prozent. Noch wichtiger ist die Wahrnehmung. Beim Vertrauen (71 Prozent), bei der Produktqualität (67 Prozent) und bei der Sortimentsbreite (66 Prozent) bewerten Verbraucher Handels- und Herstellermarken laut Ipsos und Lebensmittel Zeitung inzwischen gleichwertig. Stephan Fetsch von KPMG spricht von einer „Emanzipation der Handelsmarken als Marken sui generis“. Für D2C-Marken heißt das: Ihr Wettbewerber ist nicht nur der Billiganbieter aus Fernost, sondern auch der Händler, der Ihre Kategorie kennt, Ihre Abverkaufsdaten sieht und schneller auf Trends reagiert als Sie. Qualität allein reicht als Argument nicht mehr. Sie brauchen etwas, das die Handelsmarke nicht kopieren kann, also Community, Herkunft, echte Innovation oder eine direkte Beziehung zum Kunden. Wie teilen sich Marktplatz und eigener Shop die Arbeit? Marktplatz Eigener Shop Aufgabe Neukunden, Reichweite Marge, Bindung, Markenerlebnis Sortiment Bestseller, Einstiegsprodukte Exklusives, Bundles, Großpackungen Daten wenig Kundendaten volle eigene Kundendaten Werbung Sponsored Products, Sponsored Brands E-Mail, CRM, Community Erlösmodell Einzelkauf Abo, Nachkauf, Treueprogramm Welcher Kanal hat bei Ihnen Vorrang? Das hängt am Sortiment. Wenn Ihre Produkte austauschbar und stark preisgetrieben sind, gehört der Schwerpunkt auf den Marktplatz, der eigene Shop ist dann vor allem Schaufenster. Wenn Ihre Produkte regelmäßig nachgekauft werden, lohnt sich der eigene Shop mit Abo besonders, online wachsen Drogerie mit 11,7 Prozent und Tierbedarf mit 9,8 Prozent. Wenn Ihre Produkte erklärungsbedürftig oder personalisierbar sind, verkauft der eigene Shop besser. Und wenn Sie in der Mode gegen asiatische Plattformen antreten, gewinnen Sie nicht über den Preis, sondern über Qualität, Nachhaltigkeit oder Service. Wie rechne ich aus, welcher Kanal sich lohnt? Mit dem Deckungsbeitrag pro Bestellung nach allen Kosten des Kanals. Also Nettoumsatz minus Wareneinsatz, Marktplatzgebühren, Werbung, Fulfillment und Retouren. Klingt banal, aber ich sehe erstaunlich oft Unternehmen, die den Marktplatz nur nach Umsatz bewerten und sich wundern, warum am Ende nichts übrig bleibt. Für die Werbung auf Marktplätzen schaue ich auf den TACoS, also die Werbekosten im Verhältnis zum gesamten Umsatz inklusive der organischen Verkäufe. Der zeigt, ob Werbung auch den organischen Umsatz anschiebt. Im eigenen Shop gehört der Kundenwert über die ganze Lebensdauer dazu. Ein Abo-Kunde, der zwölfmal nachkauft, rechtfertigt ganz andere Akquisekosten als ein Einmalkäufer. Was müssen Marken auf Marktplätzen besser machen? Die Sichtbarkeit auf Amazon und Co. folgt eigenen Regeln, und die werden immer semantischer. Titel und Bulletpoints sollten Kaufkriterien und Nutzungssituationen beantworten, statt Keywords aneinanderzureihen. Die Attribute müssen vollständig gepflegt sein, weil Such- und KI-Assistenten danach filtern. A+-Content und Brand Store bauen die Marke auf. Und die Produktdaten sollten überall identisch sein, im Shop, auf dem Marktplatz und im Google Merchant Center. Wie wird der eigene Shop zum Bindungskanal? Der eigene Shop braucht einen Grund, den es auf keinem Marktplatz gibt. Exklusive Produkte, Personalisierung, ein Abo mit Preisvorteil oder ein Treueprogramm. Dass sich das rechnet, zeigen die Zahlen aus dem Lebensmittelhandel: Nutzer von Treueprogrammen geben dort im Schnitt 30 Prozent mehr aus. Ich würde Budget aus teurer Neukundenwerbung bewusst in E-Mail-Automatisierung, Nachkauf-Erinnerungen und Kundenbindung umschichten. Jeder Kunde, der vom Marktplatz in den eigenen Shop wechselt, verbessert Ihre Marge dauerhaft. Mein Fazit Marktplätze machen mehr als die Hälfte des deutschen Onlinehandels aus, und der HDE erwartet für 2026 ein weiteres Wachstum des Onlinehandels um 4,3 Prozent auf 96,3 Milliarden Euro. Auf Marktplätze zu verzichten, heißt für die meisten Marken, auf den größten Teil des Marktes zu verzichten. Gleichzeitig zeigen die Herstellerversender mit 6,3 Prozent Wachstum, dass starke Marken auch direkt verkaufen. Die profitabelste Strategie ist die Arbeitsteilung: Der Marktplatz gewinnt Kunden, der eigene Shop behält sie. Und wer das Wachstum der Handelsmarken ignoriert, wird sich in zwei Jahren wundern, wo die Regalplätze geblieben sind. Häufige Fragen Lohnt sich ein eigener Onlineshop 2026 noch? Ja, wenn die Marke stark genug ist und das Sortiment Nachkäufe oder Personalisierung erlaubt. Klassische Onlineshops ohne eigene Marke wuchsen im zweiten Quartal 2026 nur um 3,8 Prozent. Warum wächst der Onlinehandel je nach Quelle um 5,1 oder 6,4 Prozent? Weil die Quellen unterschiedlich messen. Destatis erfasst die Umsätze der Unternehmen im Versand- und Onlinehandel und kommt für Q2 2026 auf plus 6,4 Prozent nominal und plus 5,4 Prozent real. Der bevh befragt 40.000 Verbraucher zu ihren Ausgaben und kommt auf plus 5,1 Prozent. Beide zeigen dasselbe: online wächst, stationär verliert real. Wie stark sind Temu, Shein und AliExpress in Deutschland? Laut bevh erzielen asiatische Plattformen 5,3 Prozent des deutschen Onlineumsatzes und mehr als 16 Prozent der Bestellungen im Online-Modehandel. Was ist ein guter TACoS? Das hängt von Marge und Kategorie ab. Wichtiger als der absolute Wert ist der Trend: Sinkt der TACoS bei steigendem Umsatz, schiebt Ihre Werbung das organische Wachstum an. Sie wollen Marktplatz und eigenen Shop so aufstellen, dass beide Geld verdienen? Wir verbinden Marktplatz-Optimierung und Advertising mit Performance- und Retention-Strategie. Strategiegespräch vereinbaren Quellen: bevh, Pressemitteilung vom 06.07.2026; KPMG Retail Sales Monitor 3/2026 (auf Basis Destatis, bevh, YouGov, Ipsos/Lebensmittel Zeitung Handelsmonitor 2026, TW-Testclub, Hello-Again-Loyalty-Report 2026, HDE)
Retail Media im Supermarkt: Warum der Point of Sale gerade zum Performance-Kanal wird
Sep 24, 2026

Axel
Zawierucha
Category:
Growth Marketing

Kurz zusammengefasst: In-Store Retail Media verbindet Werbung im Markt mit echten Kassendaten aus Loyalty-Apps. Die Infrastruktur steht: Jede zehnte Kasse im Lebensmittelhandel ist eine Self-Checkout-Kasse, 91 Prozent der LEH-Händler nutzen elektronische Regaletiketten zumindest in Teilen. Seit 2025 ist der Loyalty-Markt neu verteilt, Edeka ist bei Payback, Rewe setzt auf die eigene App. Für Marken wird der Supermarkt damit messbar, aber nur, wer auf echte Mehrverkäufe schaut statt auf schöne ROAS-Zahlen, investiert sein Budget richtig. Wenn ich mit Markenherstellern über Retail Media spreche, geht es fast immer um Amazon. Der Supermarkt um die Ecke taucht in diesen Gesprächen kaum auf, da gibt es Handzettel, Zweitplatzierung und Jahresgespräche. Ich halte das für einen Fehler. Die Zahlen aus dem aktuellen KPMG Retail Sales Monitor zeigen ziemlich deutlich, dass sich im Lebensmittelhandel gerade ein eigener Werbemarkt aufbaut. Was ist In-Store Retail Media? In-Store Retail Media ist Werbung, die ein Händler auf seinen eigenen Flächen im Markt verkauft und mit seinen Kundendaten aussteuert. Das sind Screens im Markt, Displays an der SB-Kasse, digitale Regaletiketten und vor allem personalisierte Coupons in der Händler-App. Der Unterschied zur klassischen POS-Werbung liegt in der Messbarkeit. Bei einem Deckenhänger wusste niemand, ob er etwas verkauft hat. Wenn der Kunde dagegen an der Kasse seine App scannt, lässt sich die Werbung mit dem tatsächlichen Kauf verbinden. Damit funktioniert der Supermarkt plötzlich ähnlich wie ein Onlineshop. Warum investieren die Händler jetzt so stark? Weil das Kerngeschäft unter Druck steht und Werbeerlöse deutlich höhere Margen haben als Joghurt. Der stationäre Einzelhandel wuchs im zweiten Quartal 2026 nominal um 0,3 Prozent und verlor real 1,1 Prozent. In der HDE-Umfrage meldeten 69 Prozent der Handelsunternehmen rückläufige Gewinne im ersten Halbjahr. Gleichzeitig wird um den Platz auf dem Smartphone der Kunden gekämpft, und das sieht man an den Werbebudgets. Rewe war 2025 mit knapp 620 Millionen Euro Bruttowerbeinvestitionen der größte werbungtreibende Händler Deutschlands, ein Plus von 21,2 Prozent. Dahinter folgen Lidl mit 527 Millionen und Kaufland mit 400,2 Millionen Euro. Allein im Januar 2025, zum Start von Rewe Bonus, stiegen die Rewe-Ausgaben um 114,2 Prozent. Wer so viel Geld in Kundenbindung steckt, will es über kurz oder lang über Werbung von Herstellern zurückholen. Wie weit ist die Technik im Markt? Weiter, als viele denken. Die Zahlen dazu: Touchpoint Verbreitung SB-Kassen im LEH rund 25.000, jede zehnte Kasse im LEH (10,2 %) SB-Kassen im gesamten Einzelhandel 38.650, jede 18. Kasse Wachstum SB-Kassen im LEH von 9.600 (2023) auf 25.000 (2025) Elektronische Regaletiketten im LEH 91 % der befragten Händler (57 % in allen Filialen, 24 % in ausgewählten, 10 % im Pilot) Quellen: EHI Retail Institute, Stand August 2025; EHI Technologie-Trends 2025 Bei den digitalen Regaletiketten muss man allerdings ehrlich sein. Sie werden oft als das große neue Werbemedium verkauft. Tatsächlich zeigen sie laut KPMG noch überwiegend einfach den Preis an, Zusatzfunktionen werden kaum genutzt. Wohin es gehen kann, sieht man bei Co-op in Großbritannien: Kunden halten ihr Smartphone ans Etikett und landen direkt in der App mit dem Mitgliedspreis. Rewe nutzt die LEDs der Etiketten bisher vor allem, damit Mitarbeitende beim Kommissionieren von Abholbestellungen schneller die richtigen Artikel finden. Als buchbares Werbeformat ist das Regaletikett also noch Zukunftsmusik. Die messbaren Hebel liegen heute bei App-Coupons und Kampagnen mit Händlerdaten. Wer hat die Kundendaten im Supermarkt? Seit Anfang 2025 gibt es zwei Lager. Edeka und Netto sind zu Payback gewechselt, Rewe und Penny haben Payback verlassen und bauen eigene Programme auf. Laut einer Befragung von Workinghead und YouGov erreicht Payback eine Teilnahmequote von 67 Prozent, die Rewe App kommt auf 32 Prozent bei 52 Prozent Bekanntheit. Ganz reibungslos läuft das bei keinem. Bei Edeka ging ein Jahr nach dem Start etwa 38 Prozent des Umsatzes über die Payback-Karte, in der Region Nord weniger als 30 Prozent. Rewe verbindet seit Juli 2025 Bonus und Bezahlen in einer App (Rewe Pay), Edeka bietet Payback Pay, Lidl hat eine Bezahlfunktion in Lidl Plus. Und dann gibt es Aldi. Aldi Nord und Aldi Süd verzichten als einzige der großen Händler bewusst auf ein Bonusprogramm und haben das im Juli 2026 zur offiziellen Markenbotschaft gemacht. Sie warnen vor einer „Zweikassengesellschaft“, in der zahlt, wer keine App hat, mehr für dasselbe Produkt. Ich finde das als Positionierung ziemlich klug. Für Marken heißt es aber ganz nüchtern: Rund 35 Milliarden Euro LEH-Umsatz bei Aldi sind über Loyalty-Daten nicht erreichbar. Warum sind Loyalty-Kunden für Marken so interessant? Weil sie mehr ausgeben. Laut Hello-Again-Loyalty-Report 2026 geben Nutzer von Treueprogrammen im Schnitt 30 Prozent mehr aus als Kunden ohne Programm. 80 Prozent der Deutschen nutzen Bonus-Apps zumindest gelegentlich, 55 Prozent würden für ein Geschäft mit Programm sogar einen Umweg fahren. Wer diese Kunden gezielt ansprechen kann, erreicht die wertvollsten Käufer im Markt. Was bringt Closed-Loop-Attribution wirklich? Closed-Loop-Attribution verbindet einen Werbekontakt über dieselbe Kunden-ID mit einem späteren Kauf. Jemand sieht eine Anzeige in der App oder auf einer Nachrichtenseite, kauft zwei Tage später im Markt, scannt seine App, und in einem Data Clean Room werden beide Ereignisse pseudonymisiert zusammengeführt. Heraus kommt ein In-Store-ROAS. Hier werde ich immer skeptisch. Ein hoher ROAS beweist nicht, dass die Werbung den Kauf ausgelöst hat. Wer ohnehin jede Woche Ihren Joghurt kauft, taucht natürlich auch im Bericht auf. Verlangen Sie deshalb von jedem Netzwerk einen Test mit Kontrollgruppe, die die Werbung nicht sieht. Erst der Vergleich zeigt, wie viel Sie wirklich zusätzlich verkauft haben. Wenn ein Netzwerk das nicht anbieten will, wäre ich vorsichtig. Wie sollten Marken einsteigen? Ich würde mit drei Dingen anfangen. Erstens Trade-Marketing und Performance-Marketing an einen Tisch holen, mit einem gemeinsamen Budget und gemeinsamen Zielen. Retail Media gehört in keine der beiden Schubladen allein. Zweitens mit einem Netzwerk und Produkten starten, die oft nachgekauft werden, dort sieht man Effekte am schnellsten. Drittens den Inkrementalitätstest schon vor Kampagnenstart vertraglich festhalten, inklusive Zugang zu den Daten. Und beim Creative bitte nicht die TV-Idee verkleinern. Am Regal oder an der SB-Kasse haben Sie zwei Sekunden. Da braucht es ein Signal: Preisvorteil, Neuheit oder ein klarer Grund, warum gerade dieses Produkt. Mein Fazit Der Lebensmittelhandel hat in den letzten zwei Jahren die Infrastruktur für messbare POS-Werbung gebaut: 25.000 SB-Kassen, breit ausgerollte digitale Etiketten und Loyalty-Programme mit Millionen Nutzern. Die Händler haben 2025 so viel in Werbung investiert wie nie. Für Marken ist das die Chance, den Supermarkt endlich wie einen Performance-Kanal zu steuern. Die Voraussetzung ist, dass sie nach echten Mehrverkäufen fragen und sich nicht mit hübschen Dashboards zufriedengeben. Häufige Fragen Was kostet In-Store Retail Media? Das hängt stark von Netzwerk, Format und Zielgruppe ab. Abgerechnet wird meist nach Reichweite, bei Coupons auch nach Einlösung. Welche Marken profitieren am meisten? FMCG-Marken mit hoher Wiederkaufrate und neue Produkte, bei denen es auf Probierkäufe ankommt. Ist Closed-Loop-Attribution datenschutzkonform? Seriöse Netzwerke arbeiten mit pseudonymisierten Daten in Data Clean Rooms auf Basis der Einwilligungen aus dem Loyalty-Programm. Lassen Sie sich das Verfahren vor dem Start dokumentieren. Kann man bei Aldi Retail Media buchen? Nicht auf Basis von Loyalty-Daten, weil Aldi bewusst kein Bonusprogramm betreibt. Sie wollen Retail Media sauber in Ihren Media-Mix bringen? Wir helfen bei der Auswahl der Netzwerke, beim Setup und bei der Messung echter Mehrverkäufe. Beratungsgespräch vereinbaren Quellen: KPMG Retail Sales Monitor 3/2026 (auf Basis EHI Retail Institute, Nielsen Media, Workinghead/YouGov, Hello-Again-Loyalty-Report 2026, HDE, Destatis); EHI Self-Checkout-Markterhebung 2025; EHI Technologie-Trends im Handel 2025
KI-Einkaufsagenten im E-Commerce: Warum über Ihr Produkt entschieden wird, bevor jemand Ihren Shop sieht
Sep 24, 2026

Axel
Zawierucha
Category:
Growth Marketing

Kurz zusammengefasst: KI-Einkaufsagenten stellen in Deutschland heute Warenkörbe zusammen, den Kauf bestätigt fast immer noch der Mensch. Knuspr ist seit Juli 2026 mit einer eigenen App in ChatGPT vertreten, 31,2 Prozent der Onlinekunden haben laut bevh schon Chatbots nach Produkten gefragt. Der Wettbewerb findet deshalb nicht im Checkout statt, sondern bei der Frage, welches Produkt der Agent überhaupt vorschlägt. Und das entscheidet sich an Ihren Produktdaten. Ich habe mir in den letzten Wochen genauer angeschaut, wie Knuspr seinen Online-Supermarkt in ChatGPT gebracht hat. Für mich ist das der spannendste Schritt im deutschen E-Commerce in diesem Jahr. Nicht, weil dort schon alle einkaufen, das tun sie nicht. Sondern weil man zum ersten Mal live sieht, wo die Kaufentscheidung künftig fällt: in einem Gespräch, lange bevor jemand eine Produktseite öffnet. Was ist ein KI-Einkaufsagent? Ein KI-Einkaufsagent ist eine Software auf Basis eines Sprachmodells, die eine Kaufabsicht in normaler Sprache entgegennimmt und daraus einen Warenkorb macht. Der Kunde sucht nicht mehr nach „Bio-Vollmilch laktosefrei 1 Liter“. Er sagt: „Ich brauche Essen für drei Tage, vier Personen, proteinreich, maximal 90 Euro.“ Den Rest erledigt der Agent. Er sucht, vergleicht, wählt aus. Für Hersteller und Händler hat das eine unbequeme Konsequenz. Eine zweite Ergebnisseite gibt es nicht. Wer in den Daten des Agenten nicht auftaucht, kommt in dieser Kaufentscheidung schlicht nicht vor. Was kann Knuspr heute schon? Knuspr zeigt zwei Stufen, und der Unterschied ist wichtig. Im eigenen Shop kann die KI-Agentin Maia den kompletten Einkauf abwickeln, von der Produktauswahl bis zur Zahlung. In der ChatGPT-App, die über einen eigenen MCP-Server angebunden ist, stellt der Assistent Rezepte, Wochenpläne und Warenkörbe zusammen, sogar aus dem Foto eines handgeschriebenen Einkaufszettels. Bestellung und Zahlung laufen dort aber weiterhin über den Knuspr-Shop. Das ist kein Zufall. Der voll agentische Kauf funktioniert bisher nur dort, wo Händler und Agent dieselbe Firma sind. Sobald ein fremdes System dazwischen sitzt, holt sich der Händler den Checkout zurück, und damit die Kundenbeziehung. Kaufen KI-Agenten in Deutschland schon selbst ein? Nein, jedenfalls nicht in der Breite. OpenAI hat den Kauf direkt in ChatGPT, Instant Checkout, im März 2026 wieder eingestellt. Googles Universal Commerce Protocol bietet einen Checkout in AI Mode und Gemini bisher nur Händlern in den USA an. Die Verbraucher sind auch noch nicht so weit. In einer Sonderbefragung des bevh unter 2.500 Onlinekunden haben 31,2 Prozent schon mal Chatbots nach Produktempfehlungen gefragt. Aber nur 12,7 Prozent würden einer Empfehlung direkt folgen, ohne selbst nochmal zu suchen. Für 68,6 Prozent ist es völlig undenkbar, einem KI-Agenten ihre Zahlungsdaten zu geben. Ich lese daraus Folgendes: Der Agent macht die Shortlist, der Mensch klickt auf „Kaufen“. Wer nicht auf dieser Shortlist steht, hat verloren, bevor der Kunde überhaupt in einen Shop schaut. Die Frage, ob KI selbst bezahlen darf, ist für Ihren Umsatz deshalb erstmal zweitrangig. Warum trifft es Routinekäufe zuerst? Weil KI-Assistenten genau die Entscheidungen übernehmen, auf die niemand Lust hat. Das Waschmittel, das Katzenfutter, der Wocheneinkauf. Genau diese Warengruppen wachsen online derzeit am stärksten: Warengruppe Online-Umsatz Q2 2026 Wachstum zum Vorjahr Medikamente 547 Mio. € +13,9 % Drogerie 974 Mio. € +11,7 % Tierbedarf 641 Mio. € +9,8 % Lebensmittel 1.155 Mio. € +8,9 % E-Commerce gesamt 21,2 Mrd. € +5,1 % Quelle: bevh, Q2 2026 Dazu kommt der Preisdruck. Die Inflation lag im Juli 2026 bei 2,8 Prozent, Kraftstoffe waren im August 27,7 Prozent teurer als ein Jahr zuvor. Wer sparen muss, vergleicht mehr, und ein Assistent, der in Sekunden das günstigste passende Produkt findet, kommt da gerade recht. Man sollte die Größenordnung trotzdem ehrlich einordnen. Der Online-Lebensmittelhandel kam 2025 auf 4,12 Milliarden Euro, das sind weniger als 2 Prozent des gesamten LEH-Umsatzes. Laut KPMG Consumer Snapshot nutzen 21 Prozent der Befragten Online-Lieferdienste, als wichtigste Gründe nennen sie Zeitersparnis und Bequemlichkeit. Das ist klein, aber es sind genau die Motive, die ein KI-Agent bedient. Deshalb schaue ich auf die Richtung, nicht auf das heutige Volumen. Wie entscheidet ein KI-Agent, welches Produkt er vorschlägt? Vereinfacht läuft es in fünf Schritten. Der Agent zerlegt zuerst die Anfrage in Bedingungen, also Budget, Personenzahl, Allergene, Ernährungsweise, Lieferzeit. Dann holt er sich über Feeds und Schnittstellen Preise und Verfügbarkeiten. Im dritten Schritt prüft er, ob die Produkteigenschaften zu den Bedingungen passen. Danach bildet er ein Ranking, und am Ende legt er einen Vorschlag in den Warenkorb. Der kritische Moment ist Schritt drei. Wenn bei Ihrem Produkt das Attribut „laktosefrei“ nicht sauber hinterlegt ist, fliegt es raus, auch wenn es perfekt passen würde. Der Agent rät nicht, er filtert. Wer nicht maschinenlesbar ist, ist für ihn unsichtbar. Welche Produktdaten braucht Ihr Sortiment? Strukturierte Daten, die wirklich stimmen Basis ist sauberes Product- und Offer-Markup nach Schema.org. Über die Pflichtfelder hinaus sind für Agenten vor allem Lieferzeit und Versandkosten (shippingDetails), Rückgabebedingungen (hasMerchantReturnPolicy), eindeutige Kennungen wie gtin und brand sowie Zertifikate über hasCertification wichtig. Alternativen und Zubehör lassen sich mit isSimilarTo und isAccessoryOrSparePartFor verknüpfen. Ein Detail, das in vielen Anleitungen falsch steht: Die Schema.org-Properties nutrition und suitableForDiet gibt es nur für Rezepte und Menüeinträge, nicht für Produkte. Eigenschaften wie „vegan“, „glutenfrei“ oder Nährwerte gehören bei Produkten in additionalProperty und zusätzlich in den Produktfeed. Die Attribute, nach denen Kunden heute fragen Schauen Sie sich an, wo der Lebensmittelmarkt wächst, dann wissen Sie, welche Attribute zählen. Der Bio-Markt erreichte 2025 einen Rekord von 18,2 Milliarden Euro, bei einzelnen Online-Lieferdiensten liegt der Bio-Anteil bei bis zu 28 Prozent. Sportler- und Fitnessnahrung legte laut NielsenIQ um 44,7 Prozent zu, Proteinpulver sogar um 63 Prozent. Und wer vegan oder vegetarisch lebt, kauft fast doppelt so oft Lebensmittel online wie andere (25 zu 14 Prozent). Das heißt ganz praktisch: Bio-Zertifikat, Proteingehalt pro Portion und „vegan“ müssen als strukturierte Daten vorliegen, nicht nur als Satz irgendwo in der Produktbeschreibung. Sonst fallen Sie genau bei den Kunden durch, die am häufigsten digital einkaufen. Gleiche Daten auf allen Kanälen Wenn im eigenen Shop 500 ml steht, auf Amazon 0,5 l mit anderer Zutatenliste und bei Open Food Facts noch die alte Rezeptur, dann weiß das Modell nicht mehr, wem es glauben soll. Es wird vorsichtig und empfiehlt im Zweifel ein anderes Produkt. Die Grundlage ist ein bereinigtes PIM-System, aus dem alle Kanäle gespeist werden. Das klingt nach Hausaufgabe, ist aber in unseren Projekten fast immer der größte Hebel. Texte, die eine Situation beantworten LLMs suchen nach Aussagen, die zu einer konkreten Frage passen. „Günstige vegane Hafermilch kaufen, bester Haferdrink online“ hilft dabei gar nicht. „Bio-Haferdrink aus deutschem Anbau, schäumt stabil bei 65 °C und flockt auch in säurehaltigem Espresso nicht aus“ dagegen schon. Der zweite Satz beantwortet die Frage eines Barista-Fans, der erste beantwortet gar nichts. Was unterscheidet GEO von klassischem SEO? Generative Engine Optimization, kurz GEO, sorgt dafür, dass KI-Systeme wie ChatGPT, Gemini, Perplexity oder Google AI Mode Ihre Inhalte als Quelle nutzen und Ihre Produkte empfehlen. GEO ersetzt SEO nicht. Viele KI-Systeme greifen für aktuelle Informationen auf Suchindizes zurück, technisches SEO bleibt also das Fundament. Klassisches SEO GEO Ziel Ranking in den Suchergebnissen Nennung in KI-Antworten Messgröße Rankings, Klicks Erwähnungen, Zitierungen, Anteil in Prompts Logik Keywords Absichten und Folgefragen Wichtigste Daten Seiteninhalte, Meta-Tags, Links Strukturierte Daten, Feeds, konsistente Fakten Was ich Shopbetreibern jetzt rate Erstens: Messen Sie, wo Sie stehen. Nehmen Sie 30 bis 50 typische Kundenfragen und prüfen Sie, ob ChatGPT, Gemini, Perplexity und Google AI Mode Ihre Produkte nennen, und wen sie stattdessen nennen. Zweitens: Gehen Sie Ihre Top-Produkte durch und schauen Sie, welche kaufentscheidenden Attribute im Feed fehlen. Drittens: Bringen Sie Shop, Marktplätze und Datenbanken auf einen Stand. Viertens: Schreiben Sie die Texte Ihrer wichtigsten Produkte so um, dass sie konkrete Situationen beantworten. Und fünftens: Wiederholen Sie die Messung monatlich. KI-Antworten ändern sich schneller als Google-Rankings. Einen eigenen MCP-Server wie Knuspr brauchen die meisten nicht. Ein vollständiger, täglich aktualisierter Produktfeed ist für 90 Prozent der Shops der richtige Anfang. Was sagt unsere GEO-Studie dazu? Für unsere GEO-Studie 2026 haben wir 240 Prompts aus zwölf Branchen in Google AI Mode und ChatGPT ausgewertet und 5.317 zitierte URLs analysiert. Das Ergebnis ist für Shopbetreiber eindeutig: Reine Produktdetailseiten reichen nicht aus. Sie machten im Google AI Mode lediglich 3,5 Prozent und bei ChatGPT 4,7 Prozent der zitierten Quellen aus. Am häufigsten griffen beide Systeme auf FAQ-, Hilfe- und How-to-Seiten zurück, gefolgt von Blogartikeln und Vergleichstabellen. Bei konkret kaufbezogenen Prompts steigt der Anteil der Produktdetailseiten zwar auf 5,3 Prozent im Google AI Mode und 7,8 Prozent bei ChatGPT. Trotzdem bevorzugen die Systeme auch hier Inhalte, die erklären, vergleichen und konkrete Entscheidungssituationen beantworten. Für Shopbetreiber bedeutet das: Der Produktfeed liefert der KI die Fakten. Die Inhalte rund um das Produkt liefern ihr die Argumente für eine Empfehlung. Mein Fazit KI-Einkaufsagenten verändern den E-Commerce früher im Kaufprozess, als die meisten denken. Nicht beim Bezahlen, sondern bei der Auswahl. In Deutschland stellen Agenten heute Warenkörbe zusammen, der Kunde bestätigt im Shop. Welche Marke dabei vorgeschlagen wird, hängt an vollständigen, konsistenten und maschinenlesbaren Produktdaten. Wer ein Routinesortiment verkauft, sollte das nicht als Zukunftsthema behandeln, sondern als Aufgabe für dieses Quartal. Häufige Fragen Kaufen KI-Agenten in Deutschland schon selbstständig ein? Nur innerhalb eigener Händler-Apps wie bei Knuspr Maia. Über ChatGPT schlagen Agenten Warenkörbe vor, den Kauf schließt der Kunde im Shop ab. Laut bevh würden nur 9 Prozent der Onlinekunden einem Agenten den autonomen Einkauf erlauben. Brauche ich für GEO einen eigenen MCP-Server? Für die meisten Shops nicht. Wichtiger sind vollständige Produktfeeds, korrektes Schema.org-Markup und einheitliche Daten auf allen Kanälen. Wie messe ich meine Sichtbarkeit in KI-Antworten? Mit einem festen Set typischer Kundenfragen, das Sie regelmäßig in mehreren KI-Systemen abfragen. Entscheidend sind Nennungen, Position und die zitierten Quellen im Vergleich zum Wettbewerb. Ersetzt GEO die klassische Suchmaschinenoptimierung? Nein. Viele KI-Systeme nutzen Suchindizes als Quelle. Technisches SEO bleibt die Grundlage, GEO baut darauf auf. Sie wollen wissen, ob ChatGPT und Google AI Mode Ihre Produkte empfehlen? Wir messen Ihre KI-Sichtbarkeit und zeigen Ihnen, welche Daten fehlen. Potenzialanalyse anfragen Weiterlesen: Agentic Commerce https://internetwarriors.de/blog/bedeutung-agentic-commerce , Retail Media am POS, Marktplatz oder eigener Shop? Quellen: bevh, Pressemitteilung vom 06.07.2026; KPMG Retail Sales Monitor 3/2026 (auf Basis bevh, EHI, NielsenIQ, Arbeitskreis Biomarkt); KPMG Consumer Snapshot 03/2025; Destatis, Verbraucherpreise Juli und August 2026; neuhandeln.de vom 08.07.2026; shoptechblog.com vom 25.03.2026
KI-Inhalte kennzeichnen: Was der AI Act von Marketingabteilungen verlangt
Sep 15, 2026

Nelson
Wahrmann
Category:
Artificial Intelligence

Das Wichtigste in Kürze Zwei Pflichten, zwei Adressaten: Die vieldiskutierte maschinenlesbare Markierungspflicht trifft primär die Anbieter von KI-Systemen (Art. 50 Abs. 2 KI-VO). Für Marketingabteilungen und Agenturen gilt als Betreiber die sichtbare Offenlegungspflicht nach Art. 50 Abs. 4. Geltung ohne Übergangsfrist: Die Transparenzpflichten nach Artikel 50 gelten seit dem 2. August 2026. Eine Übergangsregelung bis zum 2. Dezember 2026 betrifft lediglich Altsysteme der Anbieter hinsichtlich der maschinenlesbaren Markierung – nicht deine Pflicht als Anwender. Kontext entscheidet bei Bildern: Bild-, Video- und Audioinhalte müssen offengelegt werden, wenn sie bestehende Personen, Objekte, Orte oder Ereignisse darstellen bzw. ihnen ähneln und fälschlicherweise als authentisch wirken können. Bei rein fiktiven oder stilisierten Grafiken greift die Pflicht in der Regel nicht. Redaktionsausnahme bei Texten: Typische Werbetexte, Produktbeschreibungen und Social-Media-Posts unterliegen nicht der Text-Offenlegungspflicht. Bei Texten zu Angelegenheiten von öffentlichem Interesse entfällt die Pflicht, sobald eine qualifizierte menschliche Prüfung und inhaltliche Verantwortungsübernahme stattgefunden hat. Reales Risiko: Teure EU-Bußgelder von bis zu 15 Millionen Euro sind für Marketingverstöße unwahrscheinlich. Wesentlich akuter sind wettbewerbsrechtliche Abmahnungen nach dem UWG sowie Ablehnungen und Kontensperren im Google Merchant Center oder bei Google Ads. Seit dem 2. August 2026 sind die Transparenzvorgaben aus Artikel 50 der europäischen KI-Verordnung (Verordnung EU 2024/1689) verbindlich anwendbar. Im Marketingalltag prallen seither zwei extreme Haltungen aufeinander: Die einen verfallen in Aktionismus und wollen jeden KI-unterstützten Satz mit Warnhinweisen versehen; die anderen wiegen sich in Scheinsicherheit, weil sie die Pflichten fälschlicherweise allein bei Anbietern wie Google, OpenAI oder Anthropic verorten. Beide Positionen gehen an der operativen Praxis vorbei. Als Agentur für datengetriebenes Online-Marketing begleiten wir tagtäglich Freigabeprozesse, Performance-Kampagnen und E-Commerce-Feeds. Aus dieser operativen Perspektive übersetzen wir Artikel 50 in klare, rechtssicher handhabbare Leitplanken für Marketer, E-Commerce-Verantwortliche und Content-Teams. Hinweis zur Einordnung: Dieser Beitrag stellt eine strategische und fachliche Orientierung dar und ersetzt keine Rechtsberatung. Für komplexe Kampagnenszenarien oder unklare Bildrechte empfiehlt sich die Abstimmung mit einer spezialisierten Kanzlei. Die Systematik von Artikel 50: Anbieter vs. Betreiber Das grundlegende Missverständnis rührt daher, dass Artikel 50 der KI-Verordnung zwei vollkommen unterschiedliche Pflichten an zwei unterschiedliche Rollen adressiert. Wer die Begriffe „Anbieter“ (Provider) und „Betreiber“ (Deployer) verwechselt, zieht zwangsläufig falsche Schlüsse für das eigene Risikomanagement. Kriterium Artikel 50 Absatz 2 (Anbieter) Artikel 50 Absatz 4 (Betreiber) Wer ist betroffen? Entwickler und Bereitsteller von KI-Systemen (z. B. Google, OpenAI, Anthropic, Midjourney). Anwenderunternehmen, Marken, E-Commerce-Händler und Agenturen, die KI-Systeme einsetzen. Was wird verlangt? Maschinenlesbare Markierung der Ausgaben (z. B. SynthID, C2PA-Metadaten, statistische Text-Wasserzeichen). Wahrnehmbare Offenlegung gegenüber den Personen, die mit dem Inhalt in Berührung kommen. Geltungsbeginn & Fristen Gilt seit 2. August 2026. Für Altsysteme (vor dem 2. August 2026 im Markt) gilt eine Übergangsfrist bis zum 2. Dezember 2026. Gilt unmittelbar seit dem 2. August 2026 – vollständig ohne Schon- oder Übergangsfrist. Einfluss im Marketing Passiert serverseitig beim Modellanbieter. Kann vom Anwender weder manuell gesetzt noch entfernt werden. Erfordert operative Kennzeichnungsprozesse bei relevanten Bild-, Audio-, Video- und Fachtext-Inhalten. Wichtig für Marketingleiter: Die häufig zitierte Übergangsfrist bis zum 2. Dezember 2026 schützt dein Unternehmen nicht vor Offenlegungspflichten. Sie räumt den Softwareherstellern lediglich Zeit ein, ihre technischen Markierungsverfahren bei bereits veröffentlichten Modellen nachzurüsten. Die Pflicht für Betreiber, täuschende Medien offenzulegen, läuft längst. Rollenverteilung: Wer haftet bei Agenturarbeit? In der Praxis entstehen Werbemittel selten isoliert in einer Abteilung. Häufig konzipiert eine externe Kreativ- oder Performance-Agentur die Kampagne, während der Kunde das Budget freigibt und die Kampagnen über eigene oder Agentur-Accounts schaltet. Wer gilt hier als Betreiber? Nach den Leitlinien der Europäischen Kommission kommt es maßgeblich darauf an, wer das KI-System unter eigener Verantwortung einsetzt und wer über den Einsatz der Technologie sowie die Verwendung der Ergebnisse maßgeblich entscheidet. Die reine Schaltung einer Anzeige oder der Rechnungsempfang sind nicht allein ausschlaggebend: Eigenständiger Agentureinsatz: Erhält die Agentur ein offenes Kreativbriefing und entscheidet intern eigenständig über den Einsatz generativer Bild-Tools, trägt sie als Betreiberin primär die operative Prüf- und Kennzeichnungspflicht für die erstellten Assets. Vorgabe durch den Auftraggeber: Gibt der Kunde den Einsatz konkreter generativer Tools verbindlich vor oder liefert eigene KI-Assets an, liegt die Betreiberverantwortung primär auf Kundenseite. Kollaborativer Workflow: Entwickeln beide Parteien die Kampagne partnerschaftlich, können beide Rollen Pflichten nach sich ziehen. Gesetzliche Betreiberrollen lassen sich durch Verträge zwar nicht einseitig gegenüber Behörden abbedingen, Verträge können aber Zuständigkeiten, Freigabeprozesse und Haftungsfreistellungen im Innenverhältnis verbindlich regeln. Praxis-Musterklausel für Angebote und Verträge: „Soweit im Rahmen der Leistungserbringung generative KI-Systeme zur Erstellung von Bild-, Video-, Audio- oder Textinhalten eingesetzt werden, stellt [Auftragnehmer/Auftraggeber] sicher, dass sämtliche gesetzlichen Kennzeichnungs- und Prüfpflichten gemäß Art. 50 KI-VO erfüllt werden. Freigaben kennzeichnungspflichtiger Werbemittel bedürfen vor Veröffentlichung der schriftlichen Bestätigung durch [Auftraggeber].“ Die materiellen Pflichten nach Absatz 4: Bilder vs. Texte Artikel 50 Absatz 4 spaltet sich in zwei grundverschiedene Tatbestände auf. Während bei audiovisuellen Medien strenge Vorgaben gelten, enthält die Textregelung eine weitreichende Erleichterung für professionelle Redaktionen. A. Bild-, Audio- und Videoinhalte: Die Authentizitätsfalle Nach Art. 50 Abs. 4 Satz 1 müssen Betreiber offenlegen, wenn Bild-, Ton- oder Videoinhalte durch KI erzeugt oder manipuliert wurden und eine „bemerkenswerte Ähnlichkeit mit bestehenden Personen, Gegenständen, Orten, Einrichtungen oder Ereignissen aufweisen und einer Person fälschlicherweise als authentisch oder wahrhaftig erscheinen würden“ (sogenannte Deepfakes). Der Gesetzgeber meint damit weit mehr als manipulierte Politikervideos. Für das Marketing ergeben sich daraus drei Prüfkriterien: Keine Redaktionsausnahme: Bei realistisch anmutenden Bildern entfällt die Kennzeichnungspflicht nicht dadurch, dass ein Art Director oder Marketingleiter das Bild geprüft und freigegeben hat. Das Bild bleibt bei Täuschungsgefahr offenlegungspflichtig. Synthetische Werbegesichter: Wer vollsynthetische Models via Midjourney oder Flux generiert, die täuschend echt wie echte Kunden in einer Alltagsszene wirken, erzeugt beim Durchschnittsbetrachter den Eindruck einer realen Fotografie. Solche Motive sollten im Zweifel offengelegt werden, um UWG-Vorwürfen wegen Irreführung vorzubeugen. Kunst- und Satire-Ausnahme läuft im Marketing leer: Art. 50 Abs. 4 Satz 2 sieht zwar eine abgemilderte Kennzeichnung für künstlerische, satirische oder fiktionale Werke vor. Die EU-Kommission und Aufsichtsbehörden stellen jedoch klar, dass reine kommerzielle Werbekampagnen diese Ausnahme in der Praxis kaum beanspruchen können. B. Textinhalte: Öffentliches Interesse und Redaktionsprivileg Für geschriebene Texte ist die Hürde für eine gesetzliche Pflicht ungleich höher. Die Offenlegungspflicht greift überhaupt nur dann, wenn ein Text veröffentlicht wird, um die Öffentlichkeit über eine Angelegenheit von öffentlichem Interesse zu informieren (z. B. Gesundheit, Politik, Recht, gesellschaftliche Debatten). Typische Marketingtexte – von Produktbeschreibungen über Leistungsseiten bis hin zu Social-Media-Captions – verfolgen einen primär kommerziellen Werbezweck. Sie erfüllen das Tatbestandsmerkmal der Information über Angelegenheiten von öffentlichem Interesse in der Regel nicht. Selbst wenn ein Unternehmen zu einem gesellschaftlich relevanten Thema publiziert (z. B. Nachhaltigkeitsberichte oder Fachbeiträge), greift die entscheidende Rückausnahme: Wurde der Text einem Prozess der menschlichen Überprüfung oder redaktionellen Kontrolle unterzogen und trägt eine natürliche oder juristische Person die redaktionelle Verantwortung, entfällt die Kennzeichnungspflicht vollständig. Praxismatrix: 9 typische Fälle im Marketingalltag Wie ordnen sich konkrete Marketing-Maßnahmen in dieses Gefüge ein? Die folgende Matrix gibt eine belastbare Orientierung für das Tagesgeschäft: Marketing-Szenario Pflicht? Fachliche Begründung & Einordnung Fotorealistisches KI-Bild in Social- oder Display-Ad Ja Wirkt auf den Betrachter wie eine reale Fotografie. Keine Redaktionsausnahme bei authentisch anmutenden Medien. Synthetisches Bild eines real existierenden Produkts Ja Täuschungsgefahr über die reale Beschaffenheit; zusätzlich hohes Risiko von UWG-Abmahnungen bei abweichenden Merkmalen. Erkennbar stilisierte Illustration / 3D-Icon / Cartoon Nein Keine Authentizitätstäuschung. Kein vernünftiger Betrachter hält eine Comic-Grafik für ein authentisches Dokumentarbild. KI-generierte Produkttexte, Google Ads Copy, Captions Nein Reine kommerzielle Kommunikation, keine Information der Öffentlichkeit über Angelegenheiten von öffentlichem Interesse. Fachlicher Blogbeitrag (redaktionell geprüft) Nein Selbst bei gesellschaftlicher Relevanz greift die Redaktionsausnahme, sofern ein Mensch Fakten geprüft und freigegeben hat. Ungeprüfter KI-Text zu Gesundheit oder Regulierung Ja Thema von öffentlichem Interesse ohne menschliche Kontrollinstanz; fällt direkt unter die Kennzeichnungspflicht nach Abs. 4. Interne Kundenpräsentation, Angebot, Pitch-Deck Nein Keine öffentliche Zugänglichmachung. Absatz 4 setzt eine Verbreitung gegenüber der Öffentlichkeit voraus. Keynote / Öffentlicher Vortrag mit KI-Bildern Differenziert Folie/Vortrag an sich frei; fotorealistische Medien auf den Slides müssen jedoch als synthetisch gekennzeichnet werden. Kundenservice-Chatbot auf der Website Ja Greift über Art. 50 Abs. 1: Nutzer müssen bei der ersten Interaktion wissen, dass sie mit einem KI-System interagieren. Wie und wo kennzeichnen? Best Practice für Ads und Web Der AI Act verlangt eine „klare und unterscheidbare“ Offenlegung spätestens im Moment des ersten Kontakts. Das Gesetz schreibt jedoch keine starre Formulierung und kein festes Grafiksymbol vor. Ein häufiger Irrtum besteht in der Annahme, man müsse zwingend einen auffälligen Schriftzug direkt in jedes Werbebild hineinrendern. In der Praxis hängt die Lösung vom Ausspielungskanal ab: Plattform-Labels in Google Ads & Social Ads nutzen: Google stellt in Google Ads, Display & Video 360 und im Campaign Manager Kennzeichnungsoptionen bereit. Wer diese Häkchen setzt, erfüllt die Informationspflicht, ohne gegen Googles eigene Richtlinien bezüglich störender Text-Overlays auf Werbebildern zu verstoßen. Auch Meta bietet entsprechende Dropdowns im Ad Manager. Eingebettetes Label im Creative: Wird ein Bild außerhalb geschlossener Werbenetzwerke genutzt (z. B. direkt auf Landingpages, in herunterladbaren Whitepapern oder Pressegrafiken), empfiehlt sich ein diskretes, aber lesbares Label in der Bildecke (z. B. „KI-generiert“ oder „Synthetische Darstellung“). Das verhindert, dass der Hinweis bei Weiterverbreitung oder Screenshots verloren geht. Formulierungen: Bewährt haben sich kurze, unmissverständliche Begriffe wie „KI-generiert“ bei vollsynthetischen Medien oder „Mit generativer KI bearbeitet“ bei partiellen Modifikationen. E-Commerce & Google Merchant Center: Eigene Spielregeln Online-Händler müssen sorgfältig zwischen den gesetzlichen Pflichten des AI Acts und den separaten Richtlinien von Plattformbetreibern unterscheiden. Google hat seine Produktdatenanforderungen unabhängig von der EU-Verordnung strukturiert und verlangt bei KI-Inhalten höchste Datenpräzision. IPTC-Metadaten sind längst Standard: Google verlangt bereits seit Februar 2024, dass synthetische oder veränderte Produktbilder die IPTC-Eigenschaft DigitalSourceType (z. B. trainedAlgorithmicMedia) in den Dateimetadaten transportieren. Wer KI-Bilder ohne Metadaten oder gar mit herausgefilterten EXIF-Daten einspeist, riskiert Ablehnungen. Strukturierte Feed-Attribute für Text: Für KI-generierte Titel und Produktbeschreibungen hält Google im Merchant Center gesonderte Attribute bereit. Händler sollten diese Auszeichnungen nutzen, um algorithmische Prüfungen reibungslos zu durchlaufen. Kein Generalverbot, aber harte Konsistenzprüfung: Vollsynthetische Produktbilder führen nicht automatisch zur Kontosperre. Allerdings führen erhebliche Abweichungen zwischen dem Feed-Hauptbild und der tatsächlichen Landingpage, irreführende Produkteigenschaften oder unzulässige Werbetext-Overlays auf dem Bild zur sofortigen Ablehnung im Feed. Das geschäftliche Risiko sitzt im E-Commerce im Merchant Center – ein abgelehnter Feed stoppt Umsätze binnen Stunden. Redaktionsprozesse pragmatisch dokumentieren Die Redaktionsausnahme für Texte ist für Content-Marketing-Teams das wirksamste Instrument zur Vermeidung von Kennzeichnungsbannern. Sie funktioniert im Ernstfall jedoch nur, wenn die menschliche Prüfung im Zweifel plausibel nachgewiesen werden kann. Eine bloße Behauptung reicht bei behördlichen Nachfragen nicht aus. Dafür bedarf es keiner bürokratischen Prüfinstanz. Es genügt ein schlanker, dokumentierter Audit-Trail im Redaktionssystem (z. B. in WordPress, HubSpot oder Jira): Prüfende Person: Name oder Kürzel der verantwortlichen Fachredakteurin bzw. des Redakteurs. Datum & Version: Zeitpunkt der finalen redaktionellen Freigabe. Prüfschwerpunkte: Kurze Bestätigung, dass Fakten, Zahlen, rechtliche Verweise und Zitate unabhängig überprüft wurden. Freigabestatus: Explizite Übernahme der inhaltlichen Verantwortung vor Veröffentlichung. Ein einfaches Pflichtfeld im CMS („Redaktionell geprüft und freigegeben durch: [Name, Datum]“) schützt Redaktionsteams nachhaltig und dokumentiert die Einhaltung der gesetzlichen Kriterien. Reale Risiken: UWG und Kontensperren schlagen EU-Bußgelder Die im AI Act genannten Bußgelder von bis zu 15 Millionen Euro oder 3 Prozent des weltweiten Jahresumsatzes erzeugen Schlagzeilen, sind aber für kleinere Kennzeichnungsverstöße im Marketingalltag ein theoretisches Schreckgespenst. Aufsichtsbehörden richten ihren Fokus zunächst auf Hochrisiko-KI und systemische Risiken großer Tech-Plattformen. Die echten operativen Gefahren für Marketingleiter liegen woanders: Wettbewerbsrechtliche Abmahnungen (§ 5 UWG): Verfälscht ein KI-generiertes Bild die Eigenschaften eines beworbenen Produkts oder täuscht über reale Kundenbewertungen und Anwendungsfälle hinweg, können Mitbewerber oder Abmahnverbände wegen Irreführung durch Unterlassen abmahnen. Abmahnungen sind schnell zugestellt, kostenintensiv und erzwingen strafbewehrte Unterlassungserklärungen. Automatisierte Ablehnungen in Werbekonten: Plattformen wie Google oder Meta reagieren auf Richtlinienverstöße algorithmisch. Werden Feed-Bilder oder Anzeigen wegen unzulässiger Overlays oder fehlerhafter Metadaten gesperrt, steht die Performance-Maschine still. Checkliste für Marketing- und Kampagnenleiter Aktiven Medienbestand scannen: Laufende Display-, Social- und Website-Kampagnen auf fotorealistische KI-Bilder prüfen. Wo reale Personen, Produkte oder Szenen suggeriert werden, Kennzeichnungen nachrüsten. Plattform-Features aktivieren: In Google Ads, Merchant Center und Social-Plattformen die vorgesehenen KI-Häkchen und Attribute aktivieren, statt Textmanuell in Grafiken zu stempeln. Feed-Metadaten prüfen: Sicherstellen, dass das DAM- oder PIM-System IPTC-Metadaten (DigitalSourceType) bei synthetischen Produktbildern korrekt exportiert und nicht beim Web-Upload abstreift. Redaktions-Workflow absichern: Ein kurzes Freigabefeld für Fachtexte im Redaktions-Tool einrichten, um den menschlichen Prüfprozess lückenlos belegen zu können. Vertragsverhältnisse ordnen: Mit Agenturen und externen Dienstleistern schriftlich klären, wer für Kennzeichnungen und Metadaten-Compliance haftet und Freigaben dokumentiert. Häufige Fragen (FAQ) Muss jeder mit KI unterstützte Blogartikel gekennzeichnet werden? Nein. Die Pflicht greift nur bei Texten, die über Angelegenheiten von öffentlichem Interesse informieren. Zudem entfällt sie vollständig, wenn ein Mensch den Beitrag redaktionell geprüft und die inhaltliche Verantwortung übernommen hat. Reicht es, ein Label in die Bildunterschrift zu setzen? Im redaktionellen Web-Kontext kann eine klare, unmittelbar zugeordnete Bildunterschrift gültig sein. Für geteilte Social-Media-Grafiken oder Anzeigen ist sie ungeeignet, da Unterschriften beim Teilen oder Einbinden verloren gehen. Hier sind Plattform-Labels oder dezente Hinweise auf der Grafik vorzuziehen. Gilt die Übergangsfrist bis Dezember 2026 für unsere Werbekampagnen? Nein. Der Aufschub bis zum 2. Dezember 2026 gilt ausschließlich für Altsysteme der Modellanbieter hinsichtlich der maschinenlesbaren Kennzeichnung. Die Offenlegungspflicht für Betreiber gilt seit dem 2. August 2026 ohne Übergangsfrist. Fazit & nächste Schritte Der AI Act ist für Marketingabteilungen kein Innovationskiller, verlangt aber ein Ende des unbedarften Umgangs mit synthetischen Assets. Bei Texten können Content-Teams durch saubere menschliche Qualitätssicherung völlig entspannt agieren. Bei fotorealistischen Bildern und Videos hingegen müssen Kennzeichnungsprozesse und Metadaten-Workflows jetzt sauber im Kampagnenmanagement verankert sein. Möchtest du prüfen, ob deine Kampagnen, Produktdatenfeeds und Redaktionsprozesse rechtssicher und plattformkonform aufgestellt sind? Das Team von internetwarriors unterstützt dich bei der praktischen Auditierung und Optimierung deiner Content- und Werbestrategie.
ChatGPT Ads in Germany: How this new advertising channel works
Sep 15, 2026

Josephine
Treuter
Category:
Search Engine Advertising (SEA)

The most important things in a nutshell Since August 24, 2026, ChatGPT Ads have been served in Germany, as part of an expansion into 31 European markets. Since the end of August 2026, self-service access in the Ads Manager has also been open to European advertisers. Ads are only seen by users of the Free and Go plans. Plus, Pro, Business, Enterprise, and Edu remain ad-free, and accounts belonging to users under 18 do not receive ads. Instead of keywords, advertisers provide context clues. Delivery is based on the context and intent of the ongoing conversation. Advertising in a chat window was a thought experiment for a long time. Since the end of August 2026, it has become a reality in Germany: OpenAI is serving ads in ChatGPT, and companies can set up campaigns themselves in the Ads Manager. For performance teams, this is the first genuinely new advertising channel in years. The question is therefore less about whether to test it, and more about how to test it without burning budget in a still-developing system. What ChatGPT Ads are and where they appear ChatGPT Ads are paid placements within ChatGPT. They appear below a response, are marked as an ad, and are visually separated from the answer. According to OpenAI, advertising does not influence the response ChatGPT provides. An ad consists of a company name, logo or favicon, headline, ad copy, image, and landing page. We recommend 16 to 24 characters for the headline and 32 to 48 characters for the text; the headline and text should offer different benefit arguments. The decisive difference from Google Ads is not the format, but the situation. In classic search, people type two to four words. In ChatGPT, they describe their starting point, their requirements, and their concerns. The ad therefore lands right in the middle of an ongoing decision-making process. Who you reach with it—and who you don't The reach figures for ChatGPT are impressive, but they do not represent the advertising-reachable target audience. Ads are exclusively served to users of the free version and the Go plan. Anyone using Plus, Pro, Business, Enterprise, or Edu will not see any ads. This is particularly relevant for B2B: anyone using ChatGPT via a company account is excluded from the target group. In B2B, the channel is therefore more of an additional touchpoint for decision-makers doing research privately or with a free account, rather than a channel to fully cover a target audience. For B2C offers, the reachable base is significantly broader at launch. Context clues instead of keywords Exact-match keywords do not exist in ChatGPT Ads. At the ad group level, advertisers define so-called context clues, which describe what an offer does, who it helps, and in which situation it is useful. OpenAI explicitly clarifies that these clues are not rigid targeting rules. Delivery is based on conversation context and intent, ad content, and landing page. In practice, this means: a term like "running shoes" does little, while a clue that connects need and usage situation does a lot. Ad groups are therefore structured around needs, themes, or product categories, not keyword lists. Offers with different messages or landing pages should be kept separate. Additionally, campaigns can be targeted to the iOS app, the Android app, or the web version, and there is geographic targeting with granularity that varies by market. For local campaigns, it's worth checking the location options before launching. 1. Campaign Manager Overview Budget, bidding, and billing OpenAI supports three billing models: CPM for reach, CPC for traffic, and oCPC, where billing is per click but delivery is optimized for a conversion event. For CPC campaigns, OpenAI currently recommends a starting maximum bid of $3 to $5, while the minimum daily budget for Euro accounts is €15 per campaign. In addition to manual maximum bids, there is the automatic strategy "Maximize Results", which, however, does not guarantee fixed efficiency targets such as target CPA or target ROAS. Two points are important for budget planning. First, the daily budget is treated as an average over seven days: on individual days, up to twice the amount can be spent, up to a maximum of seven times the daily budget over the seven-day period. Second, the technical minimum budget is not a practical test budget. A reliable evaluation requires enough impressions, clicks, and ideally conversions. Tracking and data privacy For measurement, the OpenAI Pixel and a Conversions API are available, and both can be combined. If identical conversions are sent via both pathways, the same Event ID should be used for deduplication. OpenAI appends a click reference called "oppref" to the landing page URL on ad clicks, which the Pixel can store in a first-party cookie. Additionally, UTM parameters can be set to make the traffic visible in your web analytics tool. Reporting in the Ads Manager shows impressions, clicks, cost, CTR, average CPC and CPM, as well as conversions. Advertisers do not get access to chats, chat histories, memories, or personal data; OpenAI only provides aggregated performance data. However, anyone using Pixel, Conversions API, or Advanced Matching and transferring first-party data in the process must verify the privacy compliance and the required consent themselves. This belongs in the setup and not on the list of things to figure out later. Which industries are currently a good fit for this channel In the launch phase, OpenAI is focusing on consumer goods and lifestyle, homeware, local services, travel and experiences, as well as digital products and educational offerings. The channel is particularly interesting for e-commerce, as product feed-based campaigns are already supported, including Multi-Product Carousel and reporting at the level of individual product cards. In contrast, more is excluded than many might expect: financial and health services are currently not permitted outside the US, nor is legal advice. Individual job postings and individual real estate listings may not be advertised, though general platforms can be. Political advertising, gambling, alcohol, and tobacco are also excluded. Checking the advertising guidelines should therefore be the very first step in planning. How to start a controlled test 1. Create an Ads Manager account and set up company details, billing, payment method, and team access. 2. Check landing pages. OAI-AdsBot and OAI-SearchBot must not be blocked, and the page should lead directly to the offer, not to the homepage. 3. Structure campaigns by needs and use cases, and write context clues in natural language. 4. Prepare several clearly distinguishable creative variations per offer. 5. Set up tracking before launching media: Pixel or Conversions API, UTM parameters, and consent check. 6. Start with a dedicated learning budget and evaluate based on leads, registrations, or purchases, not clicks. Conclusion: its own channel with its own logic ChatGPT Ads are not Google Search in a chat window, nor are they a replacement for existing search, social, or shopping campaigns. They are a channel of their own, where situation and need are the planning metrics. At the same time, the system is young: personalization is missing in the European Economic Area, Germany benchmarks are also lacking, and features are added almost weekly. This is exactly where the advantage lies for early testers. Those who start now with clear use cases, clean landing pages, and robust tracking will build a learning curve before the competition discovers the channel. Want to find out if ChatGPT Ads fit your business model? We at internetwarriors support the channel from strategy to conversion tracking. Feel free to contact us. FAQ on ChatGPT Ads in Germany Since when have ChatGPT Ads been available in Germany? Ads have been served in Germany since August 24, 2026. OpenAI announced the European rollout on August 18, and self-service access followed at the end of August. Can I book ChatGPT Ads myself? Yes. The Ads Manager is available to European advertisers as a self-service tool. Campaigns can be created, managed, and analyzed there, and now even using natural language via an Ads Manager plugin in ChatGPT. How much do ChatGPT Ads cost? Billing is based on CPM, CPC, or oCPC. For CPC, OpenAI currently recommends a starting maximum bid of $3 to $5, while the minimum daily budget for Euro accounts is €15 per campaign. Germany-specific average values for CPC, CPM, or conversion costs are not yet available. Are there keywords in ChatGPT Ads? Not in the traditional sense. Advertisers provide context clues regarding topics, needs, and usage situations. These help with matching but do not guarantee delivery for specific words. Do advertisers get access to the chats? No. Chats, chat histories, memories, and personal details remain inaccessible. OpenAI only provides aggregated performance data, such as impressions and clicks.
GEO, SEO, and Google Ads: Why AI Search is Changing the Rules for Digital Visibility
Sep 1, 2026

Axel
Zawierucha
Category:
Search Engine Optimization

Google is increasingly transforming search from a list of results into an AI-powered decision environment. As a result, SEO, Generative Engine Optimization (GEO), Google Ads, and data quality are merging into a unified Search Growth System. What matters now: The facts at a glance No automatic AI visibility through SEO: Within twelve months, the overlap between traditional SEO rankings and AI-cited sources dropped from 52% to under 20% ( internetwarriors GEO Study, May 2026 ). Good rankings are no longer a free pass. Visibility outside the top 50: 70–80% of the URLs cited by AI systems do not come from Google's top 50. If you only look at classic ranking positions, you'll often remain invisible in AI answers. The shift to a decision environment: 93% of sessions in Google AI Mode end without a click on a website. The AI answer itself is becoming the place where buying decisions are made. Google Ads goes AI-native: Since May 2026, Google has been testing four Gemini-based ad formats that require structured website and product data. GEO readiness is therefore becoming a performance prerequisite for paid media. Anyone looking for a product, a service provider, or a solution today is less and less likely to get just ten blue links. Users ask more complex questions, compare options directly in search, and expect a specific, understandable recommendation. With AI Overviews, AI Mode, and new Gemini-based ad formats, Google Search is moving closer to a research, consultation, and buying environment. Google itself states that 75% of users reach a decision faster and with more confidence using AI Mode (Source: Google Marketing Live, May 2026). For businesses, this fundamentally changes the requirements for digital visibility: It is no longer enough to rank solely for individual keywords or to blindly direct ads to a landing page. Brands must be equally understandable, trustworthy, and compatible for search engines, generative AI systems, and AI-powered ad formats. In this context, GEO does not replace classic SEO. GEO extends SEO by adding a crucial dimension: visibility in AI-generated answers, source lists, recommendations, and conversational search processes. What does GEO actually mean in practice? GEO stands for Generative Engine Optimization . It refers to the targeted optimization of content, brand, and product information for generative search and response systems – such as Google AI Mode, Google AI Overviews, ChatGPT, Claude, Perplexity, Gemini, or Microsoft Copilot. While traditional search engine optimization mainly aims to make pages discoverable for relevant search queries and achieve good rankings, GEO raises a new set of core questions: Is a brand actively mentioned in an AI response? Is your own website linked or cited as a source? Does the system understand in detail what product or service a company offers? Can the AI clearly explain exactly who a product or service is suitable for? Are features, benefits, limitations, prices, and use cases represented correctly? Is a brand included in comparisons, recommendations, and specific purchasing decisions? Therefore, GEO is by no means simply "SEO with AI." It is optimization for a new search world where systems autonomously pull information from various sources to generate a tailored answer, a selection, or a recommendation. Why GEO and SEO are inevitably moving closer together It's important to understand that Google does not separate generative search entirely from its existing infrastructure. According to Google's official guidelines, AI Overviews and AI Mode are still built on the proven search systems, quality mechanisms, and the familiar index of Google Search. For complex questions, Google uses a process called Query Fan-out : The system generates multiple related search queries in parallel to research different aspects of a topic in depth, and then combines them into a coherent answer. Google also makes it clear: You don't need special AI files or new markup to appear in these features – established SEO best practices remain the key factor (Source: Google Search Central, "AI Features and Your Website", as of July 2026). We also have to consider that generative systems work on two levels: live search on the web (Retrieval-Augmented Generation / RAG) and trained model knowledge. While clean SEO structures primarily feed the RAG process in real-time, a sustainable, trustworthy digital presence shapes the AI's long-term model knowledge. The conclusion is clear: Without a solid SEO foundation, there is no reliable GEO foundation. A website must be technically accessible and indexable without errors. Content must be clearly structured, helpful, up-to-date, and original. Google explicitly recommends continuously applying existing SEO best practices rather than relying on isolated "GEO hacks." SEO remains the foundation for: Crawlability and error-free indexability Technical quality and a great Page Experience A clear page and information architecture Relevant content aligned with real search intent Internal linking and building topical authority Structured product, organizational, and local information Trustworthiness, proven expertise, and freshness GEO builds on this foundation. It expands optimization by asking whether content is written with enough precision, reliability, and nuance that a generative system can seamlessly use it as a suitable information source, a reference, or a direct recommendation. SEO and GEO: Same basis, different results Even though SEO and GEO are closely linked, they don't necessarily lead to the same results. A top position in the organic SERPs is no guarantee that a page will also be cited in AI Overviews, in AI Mode, or in external AI searches. Generative systems don't construct answers from a single ranking: They analyze sub-questions, evaluate different sources, consider entities, assess freshness, and deliver an answer with the highest possible information density. The third edition of our internetwarriors GEO Study (May 2026; 240 prompts across 12 industries, 5,317 evaluated URLs in Google AI Mode and ChatGPT) shows how much the results are now diverging: The overlap between classic SEO rankings and AI-cited sources fell from 52% to under 20% within twelve months. At the same time, 70–80% of URLs cited by AI systems do not come from Google's top 50. This means SEO and GEO visibility are measurably two different playing fields – on the same technical foundation. Feature SEO GEO Focus Optimized for stable rankings and continuous organic traffic Optimized for mentions, sources, recommendations, and AI visibility Foundation Aligned with traditional search queries and search intents Aligned with prompts, decision-oriented questions, and conversational contexts Primary Goal Getting the user to click through to your own website Relevance within the AI response and highly qualified follow-up actions KPIs / Measurement Rankings, impressions, click-through rate (CTR), and sessions Mentions, citations, AI visibility, source share, and AI referrals Environment Traditional SERPs (Search Engine Result Pages) as the central hub AI Overviews, AI Mode, chatbots, and agentic search processes The correct takeaway here is definitely not: "SEO doesn't work anymore." Instead, it is: SEO alone no longer covers the entire search visibility landscape. Google Search is becoming a decision environment This shift becomes especially clear when users stop entering short keywords and instead describe their complex scenarios. Instead of searching for "CRM agency Berlin," the question in the future is more likely to be: "Which CRM solution is suitable for a growing B2B agency in Germany that wants to combine project management, sales, GDPR, and Microsoft Teams?" A traditional search engine delivers a list of links, ads, and comparison portals. A generative search, on the other hand, tries to truly understand the requirements, research individual aspects, categorize options, and provide a clear, pre-selected shortlist. The impact is measurable: According to our GEO study (May 2026), 93% of sessions in Google AI Mode end without a click to a website . The response itself becomes the place where the decision is made – and the question is no longer just whether a brand can be found, but whether it is featured in that response at all. For businesses, this makes it critical to provide precise answers to deeper questions on their own website: Who is our exact target audience for this offering? What specific problem do we solve in the long term? What technical or organizational prerequisites apply? What integrations, features, or performance limits exist? How exactly do we differ from other market alternatives? What real examples, customer cases, reviews, or professional evidence can we show? What is the transparent cost of the offer, and how does onboarding work? The more precise, reliable, and consistent this information is across the web, the better an AI system can understand your brand and put it into the right context. From keywords to prompts and intents Keywords obviously remain highly relevant for SEO and Google Ads. They continue to represent demand and help analyze topics, search volume, and commercial intent. But the customer journey is expanding. AI search often turns a short query into a logical chain of research, comparison, and decision questions. Businesses should therefore not only build keyword clusters but also prompt and intent clusters. Traditional Keywords Generative Questions and Prompts SEO Agency Berlin Which SEO agency is right for a B2B company with complex products? Google Ads Agency How can I efficiently scale Google Ads if my tracking is incomplete? GEO Agency How does a brand become more visible in Google AI Mode and ChatGPT? SEO Audit What technical and content issues are preventing AI visibility? Performance Marketing Consulting How do I connect paid media, SEO, and CRM data for better lead quality? The difference is strategically important: Keywords usually describe only a broad topic. Generative questions, however, contain the specific situation, individual constraints, target groups, challenges, selection criteria, and the desired outcome. Content that just mindlessly repeats a keyword is simply no longer enough. What's needed is content that supports informed decisions. What truly makes content GEO-optimized GEO optimization is not about writing AI texts for AI systems. Nor does it mean creating artificial paragraphs, forcing every detail into an FAQ format, or misinterpreting a file like llms.txt as a magic trick. As a technical mapping for AI crawlers, such a file in markdown format is definitely useful – but it cannot replace substance in your content. Google explicitly advises against relying on short-term tricks for generative features. The focus should solely be on unique, valuable content that helps real people. GEO-optimized content stands out through five core qualities: 1. Clear answers to specific questions Instead of rambling broadly about a topic, great content answers clearly defined questions. For example, a B2B page shouldn't just state that it offers "holistic online marketing." It should transparently break down: What exact services does the collaboration include? For which phases and company sizes is this offer suitable? What do the typical milestones and project phases look like? What data and resources are required from the client side? 2. Verifiable facts and evidence Generative systems need information they can clearly categorize and validate. This includes reliable product data, specific service descriptions, transparent processes, expertise, studies, cases, and customer testimonials. The effect is quantifiable: A study by Princeton University and IIT Delhi shows that statistics increase visibility in generative responses by about 17%, while citations and proper sources boost it by up to 33% (Aggarwal et al., "GEO: Generative Engine Optimization", 2024). The less a website relies on generic marketing buzzwords, the better. Claims like "leading" or "customized" are only helpful if backed up by facts. 3. Clear entities and validation An AI needs to be able to identify exactly who a company is, what it offers, and how services, people, and topics are connected. Implementing structured data (Schema.org) with precise sameAs links (e.g., to official commercial registers, Wikidata, or established industry profiles) is key here. It allows the Google Knowledge Graph and generative models to validate your brand's entity beyond doubt. Details about company name, portfolio, industry focus, and expert profiles should also be kept consistent across Google Business Profile, media coverage, and partner sites. 4. Structure that people and systems understand Clear headings, logical sections, tables, comparisons, and concise summaries help users just as much as search systems. LLMs break pages down into individual segments (chunks) and prefer to cite passages that make sense even without full context. A service page shouldn't just consist of an emotional intro and a contact form; it needs to explain the offer in clear, self-contained sections. 5. Continuous freshness and maintenance Outdated information on pricing, product features, legal requirements, or technical integrations hurts your digital credibility. Freshness directly impacts citation likelihood: Content that is less than three months old or has been recently updated is cited significantly more often by AI systems. GEO is therefore not a one-off content project, but an ongoing process of monitoring, optimization, and quality control – including visible update dates and dated statements for time-sensitive topics. Why Google Ads must become more AI-native Google is not only developing organic search; ad formats are also changing fundamentally. At Google Marketing Live on May 20, 2026, Google introduced four new Gemini-based ad formats for search: Conversational Discovery Ads , Highlighted Answers , AI-powered Shopping Ads , and the Business Agent for Leads . These formats are designed to connect with users right when they are in an AI-assisted research or decision-making mindset (Source: Google, "A new generation of ads for the AI era of Search", May 20, 2026). With Conversational Discovery Ads , Gemini dynamically designs the ad creative to match a user's specific, complex question – rather than serving a static asset. Highlighted Answers allow high-quality ads to appear directly in a recommendation list generated by AI Mode; both formats carry a "Sponsored" label and are accompanied by an explanatory text independently written by Gemini. For AI-powered Shopping Ads , Gemini selects suitable products from inventory data and explains why a product fits the respective situation. With the Business Agent for Leads , users interact directly with a brand agent that answers questions based on the provided company website – instead of filling out a static lead form. As of July 2026, Conversational Discovery Ads and Highlighted Answers are being tested in the US on mobile and desktop. AI-powered Shopping Ads are expected to follow later this year, and the Business Agent for Leads is in open beta in the US. A European rollout is on the horizon – and businesses that have their data and content ready now will gain a clear head start. From landing page to knowledge base In traditional paid search logic, the website is primarily the destination after the click. In an AI-native search environment, the website becomes much more: it turns into the central knowledge base, data source, and the crucial proof of trust before the final conversion. When an AI system categorizes an ad, explains product benefits, or answers a user query in a brand dialog, it needs reliable information. Depending on the ad format, this data comes from various sources: Website content and structured service pages Detailed product and service information Complete Merchant Center feeds (pricing, availability, variations) Frequently Asked Questions (FAQ) and valid support content Google Business Profile and external reviews First-party conversion signals and CRM data A weak or inconsistent website won't technically stop your Google Ads from running. However, it will mean your ads in AI-based formats are less explainable, less relevant, and ultimately less conversion-strong. Is GEO becoming a requirement for Google Ads? The short answer is: Not as a technical requirement, but as an increasingly critical performance driver. Google won't exclude businesses from the ad auction just because their website isn't optimized for GEO. Ad accounts, campaigns, and bids remain independent factors. But the more advertising gets integrated into AI-generated answers and conversations, the more important the quality signals created by strategic GEO become. Theoretical Claim Realistic Take "Without GEO, nobody will be able to run Google Ads in the future." Overstated and technically incorrect. "A website must be cited in ChatGPT or AI Mode for Google Ads to work." Not proven. Paid systems run on their own auctions. "GEO readiness improves the foundation for AI-native paid formats." Plausible and highly strategic – the new formats run on website and product data. "Good data, structured content, and valid brand info make ads highly compatible." Directly aligned with Google's new agentic ad formats. The key difference is: GEO is not a technical ticket to enter the traditional Google Ads auction. However, GEO readiness is becoming the quality foundation for paid media in AI-guided decision processes – because Gemini can only explain and recommend what is clearly described in structured data and content. What GEO readiness means for paid media For paid teams, it will no longer be enough to only optimize campaign structures, keywords, bids, and ad copy. While those factors remain important, they are now joined by data and knowledge quality on the target page: Complete offer and product data: Google can only serve AI-powered results if they are clearly described in your feed or page. Clean product feeds, structured prices, availability, and precisely documented service packages are a must. Landing pages with real decision support: An AI-assisted ad sparks the initial interest. The landing page must take that interest further and directly answer questions about relevance, problem-solving capabilities, and how you stand out from alternatives. First-party data and conversion quality: AI-based campaigns need reliable signals. It's no longer just about clicks, but about the actual quality of the leads. This includes consent-compliant tracking, Enhanced Conversions, clean CRM integrations, and regular offline conversion imports. The better this data foundation, the more accurately automated bidding can optimize for valuable business results rather than surface-level interactions. Clear brand and trust signals: When users ask AI systems for recommendations, brands that are demonstrably trustworthy win. Technical expertise, verified author profiles, case studies, customer reviews, and digital PR mentions on high-authority sites strengthen brand perception and measurably increase the chances of an AI recommendation. A practical example: AI search and B2B lead generation A company uses Google AI Mode to find a solution: "Which agency can help us combine SEO, Google Ads, and conversion tracking for a complex B2B product?" In a traditional search environment, Google would display standard ads and organic results. The decision would mostly be made after the first click on the websites. In an AI-guided environment, the search engine also checks in the background: Which agencies demonstrably offer this combination? Is there structured expert content that validates these connections? Are target groups and project models precisely described? Are there verifiable case studies? An ad can buy initial visibility. But if the website only contains generic phrases, the AI lacks the data foundation to categorize and recommend the brand convincingly in the chat. The result: paid ads run technically, but the brand loses ground in the crucial research phase of AI search. In an environment where 93% of sessions end without a click, this phase becomes the ultimate bottleneck. The new unit: Search Growth For forward-thinking companies, it is vital to stop treating SEO, GEO, paid media, and analytics as separate silos. Instead, they are merging into an integrated overall system: the Search Growth System . Discipline Specific Function in AI Search SEO Ensures technical accessibility, clean indexing, fundamental relevance, and organic discoverability. GEO Maximizes the likelihood of mentions, accurate citations, and direct recommendations in generative answers. Content Provides concrete, deep, and verifiable answers to complex human decision-making questions. Paid Media Activates, tests, and scales demand in commercially relevant moments. Digital PR & Brand Strengthens digital reputation, personal expertise, and external trust signals trained into the model. Data & Measurement Makes actual impact, data quality, and commercial performance measurable. CRO Seamlessly translates generated visibility and initial interest into qualified leads, sales, or inquiries. This holistic approach is a major competitive advantage, especially in B2B environments with long, consultative buying cycles and complex requirements. How businesses make GEO measurable GEO shouldn't rely on random screenshots or isolated prompt tests. Generative answers vary depending on timing, location, language model, and user context. Reliable measurement requires continuous, data-driven analysis. A professional monitoring framework – like the one used in our GEO study (240 prompts, 12 industries, three waves of data collection) – combines the following levels: Share of relevant focus prompts where your own brand is mentioned Share of prompts where the brand is actively recommended Share of AI answers with a directly linked or cited company source Analysis of mentioned competitors and the tone of the brand categorization Development of AI referral traffic and its specific conversion quality Growth of targeted brand searches Organic visibility and AI feature impressions in Google Search Console On Google's side, measurability is getting concrete: Since June 3, 2026, Google has been providing dedicated Search Generative AI Performance Reports in Search Console. For the first time, they show separately how often URLs appeared in AI Overviews, AI Mode, and generative Discover features – with data starting from May 18, 2026, broken down by pages, countries, and devices. Click, CTR, and query data are still missing in the initial version; the rollout is starting in phases, beginning in the UK (Source: Google Search Central Blog, June 3, 2026). This makes GEO measurable step-by-step in standard tools. The practical GEO checklist If you want to future-proof your digital presence for AI search and upcoming ad formats, you should systematically optimize these four core areas: 1. Technical foundation Can all strategically important pages be crawled and indexed without issues? Is internal linking logical, consistent, and easy to follow topically? Is structured data (Schema.org) implemented broadly, error-free, and semantically correct? 2. Content and offering Do dedicated, deep-dive pages exist for services, products, target groups, and specific use cases? Are complex questions answered precisely throughout the entire customer journey? Is all content concrete, original, free of empty marketing buzzwords, and backed by expertise? Do central pages contain numbers, studies, or facts that an AI can cite? 3. Entities and trust Are company name, products, key people, and locations named consistently across the web? Are external reviews, certifications, professional publications, and relevant media mentions actively maintained? Are sameAs links included in the source code to help AI validate entities? 4. Paid and data quality Are data feeds (e.g., Merchant Center) or service structured data complete and up to date? Is high-quality CRM or offline data feeding reliably back into campaign management? Are landing pages ready to answer complex user questions and handle real consulting scenarios? Summary and outlook: GEO doesn't replace SEO – it completes search visibility The future of digital reach is not "SEO or GEO." Nor is it "organic or paid." The crucial question is whether a brand is deeply understood, classified as credible, and made visible at the right time in an AI-guided customer journey – in a search landscape where less than 20% of AI-cited sources overlap with classic rankings, and 93% of AI Mode sessions end without a single click. SEO remains indispensable: It builds the technical and qualitative foundation for any form of digital discoverability. GEO expands this foundation to target visibility in generative answers, structured source lists, and personalized recommendations. Paid media ensures that companies can actively scale demand in critical moments. Tracking and first-party data make it measurable which levers actually drive business success. If you're still planning SEO, GEO, and Google Ads in strict isolation, you're missing out on a massive opportunity. Those who understand and manage them as an interconnected Search Growth System build the ultimate foundation for sustainable visibility, trust, and measurable performance in the era of AI search.
KI-ROI 2026 im Faktencheck: Was Googles Studie zeigt und was sie verschweigt
Aug 5, 2026

Axel
Zawierucha
Category:
Artificial Intelligence

Das Wichtigste in Kürze: Google Cloud meldet in der Studie "From token-maxxing to ROI" (2026), dass 84 % der Unternehmen steigende finanzielle Erträge aus KI sehen. Unabhängige Analysen widersprechen: Laut McKinsey erreichen nur 6 % der Unternehmen einen EBIT-Beitrag von mehr als 5 % durch KI, laut MIT liefern 95 % der KI-Piloten keinen messbaren Gewinneffekt. Ich habe die neue Google-Cloud-Studie am Wochenende komplett gelesen, alle Seiten, inklusive Fußnoten. Und ich sage es direkt: Die Zahlen sind beeindruckend, die Fallbeispiele sind stark, und trotzdem hatte ich beim Lesen dieses Gefühl, das ich aus 25 Jahren Agenturgeschäft kenne. Wenn eine Studie zu glatt aussieht, lohnt sich der Blick auf den Absender. Der Absender ist hier Google Cloud, also einer der größten Verkäufer von genau der Infrastruktur, deren Erfolg die Studie belegt. Das macht die Studie nicht wertlos, im Gegenteil, sie ist eine der besten Quellen für die Denkweise von Entscheidern weltweit. Aber sie erzählt nur die halbe Geschichte. Die andere Hälfte steht bei McKinsey, Gartner, IBM und BCG, und die klingt deutlich nüchterner. Beides zusammen ergibt erst das Bild, das Sie für Ihre Budgetentscheidung brauchen. Was verspricht die Google-Cloud-Studie 2026? Die Studie "From token-maxxing to ROI: A reality check for enterprise AI" basiert auf einer Befragung von 2.403 Führungskräften weltweit, durchgeführt mit der National Research Group, ausgewertet wurden 297.782 Datenpunkte. Die Kernbotschaft: Die Experimentierphase ist vorbei, KI liefert messbaren Return on Investment. Die zentralen Zahlen aus dem Report: 86 % der befragten Führungskräfte stimmen zu, dass KI kosteneffizientes Wachstum ermöglicht, also Umsatz oder Output steigert, ohne dass die Betriebskosten proportional mitwachsen. 84 % berichten von stetig steigenden oder sich beschleunigenden finanziellen Erträgen aus ihren KI-Initiativen. Die obersten 26 % mit beschleunigten Erträgen nennt Google "AI ROI Leaders". 94 % geben an, dass KI-Agenten sowohl zu Kosteneinsparungen als auch zu Umsatzwachstum beitragen. 97 % planen, ihr KI-Budget im kommenden Jahr weiter zu erhöhen. 55 % nennen schnellere strategische Entscheidungen als messbaren Effekt, damit löst dieser Wert die reine Produktivität als wichtigste Erfolgsgröße ab. Fallbeispiele von Best Buy, PayPal, Tata Steel und Coca-Cola untermauern das Narrativ, alle umgesetzt auf Google-Plattformen wie Gemini Enterprise. Und wer bis zur letzten Seite liest, findet dort den entscheidenden Hinweis auf die Motivation des Verfassers. Der Report endet mit einem Button: "Contact Sales". Warum rechnet Google anders als McKinsey, Gartner und IBM? Die Diskrepanz zwischen der Google-Studie und unabhängigen Analysen hat drei Ursachen, und keine davon ist ein Skandal. Man muss sie nur kennen, um die Zahlen richtig einzuordnen. 1. Kommerzielle Interessen des Absenders Google Cloud ist kein Forschungsinstitut, sondern verkauft Rechenleistung, Datenplattformen und KI-Modelle. Jeder KI-Agent, jede Abfrage und jeder verarbeitete Token erzeugt laufende Umsätze. Eine Studie, die Investitionsdruck bei Vorständen aufbaut und Budgetbedenken abräumt, ist in diesem Kontext ein Vertriebsinstrument. Der abschließende "Contact Sales"-Button macht daraus auch kein Geheimnis. 2. Wahrnehmung statt Bilanz Google fragt nach Zustimmung zu Aussagen, also nach der subjektiven Einschätzung von Managern. Wenn 86 % zustimmen, dass KI kosteneffizientes Wachstum ermöglicht, misst das Erwartung und Überzeugung, nicht den auditierten EBIT-Beitrag. McKinsey, Gartner und IBM arbeiten näher an der Bilanz: Sie erheben Implementierungsquoten, Projektabbrüche und den tatsächlichen Ergebnisbeitrag in der Gewinn- und Verlustrechnung. Beide Messungen sind legitim, aber sie beantworten unterschiedliche Fragen. 3. Der Scheinwerfer auf den Vorreitern Google stellt die obersten 26 % der Befragten ins Zentrum der Argumentation, die "AI ROI Leaders" mit klaren Entscheidungsstrukturen, verpflichtenden Schulungsprogrammen und tief integrierten Systemen. Das ist analytisch sauber gemacht, verschiebt aber die Wahrnehmung. Was als Marktrealität gelesen wird, ist das Best-Case-Szenario einer hochentwickelten Minderheit. Was sagen die unabhängigen Zahlen zum KI-ROI 2026? Stellt man die Google-Aussagen den Ergebnissen unabhängiger Analysten gegenüber, wird die Lücke zwischen Marketing-Narrativ und Unternehmensrealität deutlich sichtbar: Kriterium Google Cloud Studie 2026 Unabhängige Analysen 2025/2026 Kernthese KI ist in Produktion angekommen, 86 % der Führungskräfte sehen kosteneffizientes Wachstum Große Lücke zwischen Nutzung und Ertrag, rund zwei Drittel kommen über die Pilotphase nicht hinaus (McKinsey) ROI-Quote 84 % melden stetig steigende oder sich beschleunigende Erträge Nur 6 % erreichen einen EBIT-Beitrag von über 5 % (McKinsey), nur 25 % der Initiativen liefern den erwarteten ROI (IBM) KI-Agenten 94 % schreiben Agenten Kosteneinsparungen und Umsatzbeiträge zu Nur 16 % haben KI unternehmensweit skaliert (IBM), über 40 % der Agentic-Projekte werden bis Ende 2027 abgebrochen (Gartner-Prognose) Projektstatus Vom Piloten in den Regelbetrieb, Fokus auf Entscheidungstempo 95 % der Piloten ohne messbaren Gewinneffekt (MIT), 42 % der Unternehmen brachen 2025 den Großteil ihrer KI-Projekte ab (S&P Global) Messbarkeit Subjektive Einschätzung der Befragten genügt als Beleg Nur 29 % können ihren KI-ROI verlässlich beziffern (IBM), nur 28 % der Use Cases in IT-Betrieb erfüllen die ROI-Erwartung voll (Gartner) Kosten "Token-Effizienz" als gelöstes Thema Agenten brauchen 5 bis 30 Mal mehr Tokens pro Aufgabe, mindestens 50 % der GenAI-Projekte überziehen bis 2028 ihr Budget (Gartner) Die Piloten-Falle ist real Während Google den flächendeckenden Produktivbetrieb reklamiert, zeigt McKinseys State of AI Survey vom November 2025 ein anderes Bild: 88 % der Unternehmen nutzen KI in mindestens einer Funktion, aber rund zwei Drittel haben die Skalierung über Pilot- und Experimentierphasen hinaus nicht geschafft. Das MIT kam 2025 zu dem viel zitierten Befund, dass 95 % der GenAI-Piloten keinen messbaren Effekt auf die Gewinn- und Verlustrechnung liefern. Und S&P Global stellte fest, dass 42 % der Unternehmen 2025 den Großteil ihrer KI-Projekte abgebrochen haben, mehr als doppelt so viele wie im Jahr davor. Das Märchen vom breiten EBIT-Beitrag Google meldet, dass 84 % der Unternehmen steigende Erträge sehen. McKinsey misst dagegen: Nur 39 % der Unternehmen können überhaupt einen EBIT-Effekt auf Unternehmensebene ihrer KI zuordnen, und bei den meisten davon liegt er unter 5 %. Gerade einmal 6 % qualifizieren sich als High Performer mit einem EBIT-Beitrag von mehr als 5 %. Die IBM CEO Study unter 2.000 CEOs bestätigt das Muster von der anderen Seite: Nur 25 % der KI-Initiativen haben den erwarteten ROI geliefert, nur 16 % sind unternehmensweit skaliert. KI-Agenten: hohe Erwartung, frühe Lernkurve 94 % der Google-Befragten schreiben Agenten finanzielle Beiträge zu. Gartner prognostizierte dagegen schon im Juni 2025, dass über 40 % der Agentic-AI-Projekte bis Ende 2027 abgebrochen werden, wegen eskalierender Kosten, unklarem Geschäftswert oder unzureichender Risikokontrollen. Viele als Agenten vermarktete Produkte sind laut Gartner zudem kaum mehr als klassische Bots, das Phänomen hat sogar einen Namen bekommen: Agent Washing. Die Technologie ist prominent, aber die breite Masse steht am Anfang der Lernkurve, nicht am Ende. Die Token-Kostenfalle Google titelt "From token-maxxing to ROI" und erklärt Token-Effizienz zur gelösten Plattformfrage. Die Gartner-Daten zeigen das Gegenteil: Agentische Modelle benötigen 5 bis 30 Mal mehr Tokens pro Aufgabe als ein klassischer GenAI-Chatbot, und bis 2028 werden mindestens 50 % der GenAI-Projekte ihre budgetierten Kosten überschreiten, vor allem wegen schlechter Architekturentscheidungen und fehlender Betriebserfahrung. Gartner empfiehlt Entscheidern inzwischen, den ROI beim Vierfachen der heutigen Kosten zu modellieren. Der laute Ruf nach ROI ist in vielen Unternehmen deshalb kein Erfolgszeichen, sondern ein Hilferuf der CFOs nach Kostenkontrolle. Worauf sollten Entscheider bei KI-Studien achten? Absender prüfen. Wer profitiert vom Ergebnis? Eine Anbieterstudie misst Vertrauen und Erwartung, eine unabhängige Analyse misst Bilanzen. Googles 86 % und McKinseys 6 % widersprechen sich nicht, sie messen schlicht Unterschiedliches. Weiche Ersatz-KPIs hinterfragen. Schnellere Entscheidungen, höhere Nutzungsraten, gefühlte Produktivität: All das sind Indikatoren, aber keine Deckungsbeiträge. Wenn der wichtigste Erfolgsnachweis einer KI-Initiative ein weicher KPI ist, ist die Ertragslage meist ungeklärt. Die Erwartungslücke ernst nehmen. Laut BCG AI Radar 2026 erwarten rund 90 % der CEOs, dass KI-Agenten noch 2026 messbaren ROI liefern, und die Hälfte der CEOs glaubt, dass ihre Jobstabilität davon abhängt, KI richtig hinzubekommen. Gleichzeitig liefern laut IBM nur 25 % der Initiativen den erwarteten Return. Diese Lücke erzeugt Druck, und Druck erzeugt schlechte Projektentscheidungen. Betriebskosten von Anfang an modellieren. KI verursacht bei jeder Nutzung Kosten, anders als klassische Software. Wer Agenten plant, sollte den Token-Verbrauch pro Geschäftsergebnis kalkulieren, nicht pro Abfrage, und Kostenlimits vor dem Go-live definieren. Wie realisieren Unternehmen den echten KI-ROI? Heißt das alles, dass KI-Investitionen sinnlos sind? Keineswegs. In einem Punkt trifft die Google-Studie die Realität exakt: Die Kluft zwischen Unternehmen, die KI strategisch verankern, und jenen, die an der Oberfläche kratzen, wächst schnell. McKinseys High Performer sind selten, aber sie existieren, und ihre Muster sind gut dokumentiert. Vier Punkte entscheiden aus unserer Erfahrung darüber, auf welcher Seite der Kluft ein Unternehmen landet: Harte Anwendungsfälle statt pauschalem Rollout. Nicht "KI für alle", sondern konkrete Flaschenhälse in Marketing, Vertrieb oder Service identifizieren und dort einen einzigen Use Case messbar produktiv machen, bevor skaliert wird. Datenbasis vor KI-Ambition. Kein Agent liefert verlässliche Ergebnisse auf einem unvollständigen CRM oder ERP. Data Governance kommt vor KI-Governance, das ist unbequem, aber nicht verhandelbar. ROI-Metriken vor dem Start festlegen. Bearbeitungszeit pro Lead, eingesparte Dienstleisterkosten, Conversion-Effekte: Wer erst nach dem Rollout misst, kann Erfolg nicht mehr von Zufall unterscheiden. Genau daran scheitert laut IBM die Mehrheit, nur 29 % können ihren KI-ROI heute verlässlich beziffern. Schulung fest in Rollen verankern. Hier hat Google recht: Die ROI Leaders unterscheiden sich messbar dadurch, dass kontinuierliche KI-Weiterbildung verpflichtender Teil der Rollenprofile ist, 38 % gegenüber 18 % im restlichen Feld. Fazit Die Google-Cloud-Studie 2026 zeigt präzise, wie die weltweite Management-Elite über KI denkt und wohin die Erwartungen laufen. Sie beschreibt Potenzial und Wunscheffekt. Die Bilanzrealität sieht 2026 anders aus: 6 % echte High Performer bei McKinsey, 25 % erfüllte ROI-Erwartungen bei IBM, 95 % wirkungslose Piloten laut MIT. Wer beide Perspektiven kennt, trifft die besseren Entscheidungen. Erfolgreiche KI-Integration braucht keinen Aktionismus, sondern harte Anwendungsfälle, saubere Daten, vorab definierte ROI-Metriken und Kostenkontrolle vom ersten Tag an. Häufige Fragen zum KI-ROI 2026 Wie viele Unternehmen erzielen 2026 tatsächlich messbaren ROI mit KI? Je nach Messmethode zwischen 6 % und 28 %. McKinsey zählt 6 % High Performer mit einem EBIT-Beitrag über 5 %, IBM berichtet, dass 25 % der KI-Initiativen den erwarteten ROI geliefert haben, und Gartner ermittelte im IT-Betrieb 28 % Use Cases, die ihre ROI-Erwartung voll erfüllen. Anbieterstudien wie die von Google Cloud kommen auf deutlich höhere Werte, weil sie subjektive Einschätzungen statt Bilanzeffekte messen. Warum scheitern so viele KI-Projekte an der Skalierung? Die häufigsten Ursachen sind mangelnde Datenqualität, fehlende Integration in bestehende Systeme, unklare Zuständigkeiten und unterschätzte Betriebskosten. Gartner prognostiziert, dass über 40 % der Agentic-AI-Projekte bis Ende 2027 abgebrochen werden, und dass mindestens 50 % der GenAI-Projekte bis 2028 ihr Budget überziehen. Das Problem liegt selten am Modell, fast immer an Daten, Prozessen und Governance. Sind Anbieterstudien wie die von Google Cloud wertlos? Nein. Sie sind hervorragende Quellen für Stimmung, Prioritäten und Best Practices der Vorreiter, und die Handlungsempfehlungen der Google-Studie zu Entscheidungsstrukturen und Schulung sind fundiert. Wertlos werden sie erst, wenn man ihre Zustimmungswerte unkritisch als Marktrealität liest. Die Faustregel: Anbieterstudien zeigen, was möglich ist, unabhängige Analysen zeigen, was üblich ist.
Rethinking touchpoints: How LLMs are taking over the first seconds of the customer journey
Jul 27, 2026

Josephine
Treuter
Category:
Artificial Intelligence

Quick Summary : LLMs like ChatGPT, Perplexity, and Google AI Mode are becoming the first touchpoint with a brand, even before the traditional click. Traditional attribution models don't capture this touchpoint: no click, no session, no entry in Google Analytics. AI answers already influence 35% to 40% of B2B buying decisions, yet they remain completely invisible in most companies' tracking. The flip side of the coin: AI referral traffic converts up to 31% better than traditional organic traffic. A potential customer is looking for new HR software. Not via Google, but directly on ChatGPT. The answer comes in seconds: three providers, briefly compared, with a recommendation. Your brand shows up—or it doesn't. And what happens next is invisible to your marketing team. This exact scenario happens millions of times every day. LLMs are taking over the first seconds of the customer journey, and most companies don't even know it's happening. This article explains why traditional attribution models break down here, why this actually offers a great opportunity for conversion quality, and how companies can build both visibility and measurability in AI systems using GEO. Because bringing these three dimensions together is exactly what has been missing in the market so far. The touchpoint nobody tracks The traditional customer journey followed a measurable logic: search query, click, website visit, tracking. Attribution models were built on this chain—first click, last click, data-driven. Every step left a footprint. This logic has developed a blind spot. According to recent analyses, around 60% of all Google searches end without a click—a zero-click customer journey that is no longer the exception. AI Overviews appear in nearly 30% of all search queries, making them an established part of the customer journey before anyone even visits a website. By August 2025, ChatGPT already recorded 800 million weekly active users, and Perplexity processed more search queries in May 2025 than in the entire year of 2024 combined. What this actually means: customers are gathering information, forming opinions, and putting together initial shortlists completely outside of your measurement points. By the time someone does click through to a website, the decision is often already half-made. The first contact with the brand happened minutes or hours ago, in an AI response that no analytics tool ever captured. Traditional Touchpoint vs. LLM Touchpoint: A Direct Comparison The following overview shows why LLM touchpoints remain structurally invisible in traditional tracking setups: Feature Traditional Touchpoint (Click) LLM Touchpoint (AI Answer) Trigger Search query on Google, click on search result or ad Prompt to ChatGPT, Perplexity, or Google AI Mode Data Footprint Session, referrer, UTM parameters None – the recommendation happens off-website Visibility in Google Analytics Fully captured (source, medium, campaign) Not captured; follow-up visits show up as direct traffic Attribution First click, last click, data-driven – all models work No model works, the trigger remains invisible Shortlist Influence After the click, on the website Before the first trackable contact, directly in the AI answer Optimization Lever SEO: rankings, snippets, CTR GEO: citability, brand authority, structured content Measurability Standard web analytics (GA4, Matomo) Specialized tools like Peec AI (mentions, sentiment, competitor analysis) The core issue in short: If it doesn't get clicked, it doesn't show up in Google Analytics. If it's recommended in an AI answer, it leaves no trace in classic attribution models. This isn't just a small data gap. This is the very beginning of the customer journey. Why the tracking model breaks down right here Recent studies show that AI search results, chatbot conversations, and voice assistant recommendations already influence 35% to 40% of B2B purchase decisions. Yet, most companies fail to track this channel because it structurally doesn't generate a referral session. The GEO measurement problem goes deeper than it seems at first glance. There are three specific friction points in traditional attribution setups: Zero-click responses don't create a referral session. A customer who asks ChatGPT and then directly visits your website appears in tracking as direct traffic, not as an AI referral. Shortlist influence happens before the first trackable touchpoint. Whether a brand even makes the shortlist is decided in the AI response, minutes before someone searches on Google or enters the URL directly. According to our GEO study , only 9.2% of sources cited by ChatGPT rank in Google's top 10 search results. This means SEO rankings and LLM visibility are largely disconnected. This means: if you only measure SEO KPIs, you miss where the journey actually begins. And if you don't see where the journey begins, you can't optimize it. The bright side: Why AI traffic converts better So far, this shift might feel like a pure loss of control. But the data shows something interesting: traffic coming from AI responses is of much higher quality than traditional organic traffic. According to Adobe Digital Insights, AI referral traffic converts up to 31% better than traditional SEO traffic. This is due to a simple mechanism: AI responses pre-qualify users in a way no search result can. Anyone arriving on a website via a recommendation on ChatGPT or Perplexity has already asked a specific question, received a structured answer, compared options, and made an initial decision. In short: less overall traffic, but highly qualified traffic. Fewer visitors who just stumbled onto your site, and more visitors with high purchase intent. B2B and B2C: Where the shift is felt most B2B: The silent shortlist In a B2B context, this shift is especially profound. Decision-makers and buying committees increasingly research anonymously and with AI support before making contact. Questions like "Which ATS software is best for mid-sized companies?" or "Which SEO agency has B2B experience?" are taken straight to ChatGPT. The AI's answer decides who makes the longlist—and who is left out. The sales cycle challenge: if you aren't present during this phase, you might never even make the evaluation. Not because your product is inferior, but because your brand isn't sufficiently anchored in the information systems that LLMs pull from. B2C: Product comparisons directly in AI In B2C, the traditional comparison process is shifting. Questions like "Which TV wall mount is best for large screens?" or "Which health insurance is best for freelancers?" are increasingly answered by LLMs with concrete recommendations, skipping price comparison sites entirely. Appearing as a brand in these responses builds relevance right at the moment of decision-making. GEO: Combining visibility and measurability This is where the major gap in the market lies: most articles about GEO either explain the problem (LLMs are changing the journey) or offer a technical solution (how to structure your content for LLMs). What is missing is the link between visibility, conversion quality, and concrete measurability. GEO (Generative Engine Optimization) is the practice of optimizing content, brand authority, and reputation signals so that they appear in AI-generated answers. GEO touchpoints are not just a nice add-on to the traditional journey; they are its new beginning. But GEO without measurement is like SEO without rankings. You just don't know if it's working. That's why a complete GEO approach combines three levels: Building Visibility: Structuring content so LLMs can easily process and cite it. Keep it factual and clear, without marketing fluff. Anchor authority through external sources, mentions, structured data, and consistent brand messaging. Understanding Conversion Quality: Analyzing AI referral traffic separately. Which prompts lead to what traffic? How does this traffic perform compared to traditional organic traffic? Measuring LLM Visibility: With tools like Peec AI, you can measure LLM visibility—how often and in what way your brand appears in answers from ChatGPT, Perplexity, and Google AI Mode, including sentiment analysis and competitor comparisons. This is the starting point for an attribution model that actually captures the very first touchpoint. The key takeaway: GEO ROI is structurally harder to measure than SEO because zero-click answers don't generate referral sessions. However, building tracking frameworks today that capture AI touchpoints gives you a data advantage that competitors will struggle to catch up with later. Conclusion: Three questions every marketing team should ask right now The customer journey is starting fresh—not on a search results page, but in a conversation with an AI. For marketing teams, this means a fundamental shift in perspective: Where does our journey actually begin? Not: "Which channel brings the first click?" but rather: "In which AI response does our brand show up before anyone even searches?" What are we missing? What part of our decision pipeline is currently invisible because it happens before the first trackable touchpoint? How qualified are our visitors when they arrive? If AI referral traffic converts better, it is highly worth identifying and specifically targeting this group. Companies investing in GEO and its measurability today are building a foundation that gets more valuable every month. Those who wait will keep optimizing for a journey that started somewhere else a long time ago. Want to know how visible your brand is today in ChatGPT, Perplexity, and Google AI Mode? Let's chat. FAQ: Common questions about LLMs, touchpoints, and GEO What is the difference between a traditional touchpoint and an LLM touchpoint? A traditional touchpoint, like a click on a search result or an ad, is visible to analytics tools and leaves a trackable data trail. An LLM touchpoint occurs when a user asks an AI like ChatGPT or Perplexity and gets a recommendation without necessarily clicking through to a website. This brand contact happens, but it is completely missed by traditional attribution models. Why does attribution get harder with LLMs? Because zero-click AI answers don't generate referral sessions. If someone visits a website directly after an AI recommendation, they show up in tracking as direct traffic. The actual trigger—the AI response—remains invisible. This makes it structurally harder to assign credit for a purchase decision to the right channel. Does AI referral traffic really convert better? Yes, according to recent analyses (Adobe Digital Insights, Seer Interactive 2025), traffic from AI recommendations converts up to 31% better than traditional organic SEO traffic. The reason: AI responses qualify users before that first click even happens. Someone coming via an AI recommendation has already asked a specific question, read a structured answer, and made an initial choice. What is GEO and why is SEO alone no longer enough? GEO stands for Generative Engine Optimization. It covers all measures companies take to ensure their content and brand show up in AI-generated responses. While SEO optimizes for rankings and clicks, GEO is about being a citable source in systems like ChatGPT, Perplexity, or Google AI Mode. Since our internetwarriors GEO Study 2026 shows that only 9.2% of sources cited by ChatGPT rank in Google's top 10, SEO rankings and LLM visibility are largely independent of each other. How can I measure LLM visibility? With specialized tools like Peec AI, you can systematically track whether and how often your brand appears in responses from ChatGPT, Perplexity, Google AI Mode, and other LLMs, alongside sentiment analysis and competitor insights. Traditional web analytics tools simply don't capture this channel. Setting up a GEO tracking framework today gives you a valuable data foundation for the future.
Performance Max Campaigns: Advanced Strategies and Pitfalls for 2026
Jul 13, 2026

Yasser
Teilab
Category:
Search Engine Advertising (SEA)

The most important details at a glance: Advanced Control 2026: Performance Max has become more transparent thanks to campaign-wide exclusions, detailed channel performance reports, and granular asset metrics, but it remains a system that needs tight guardrails. Profitability Before Algorithm: Budgets and campaign splits should not be based on purely visual categories, but on hard business metrics such as margins, product lifecycles (evergreen vs. longtail), or customer value. Signposts Instead of Targeting: Audience signals, search themes, and customer match serve as signposts for Google AI and must not be misunderstood as rigid, exact targeting. The focus must be on high-quality first-party data. From ROAS to POAS: A high ROAS often covers up unprofitable sales segments. Advertisers should establish Profit on Ad Spend (POAS) as the primary steering metric via cart data import. Hybrid Account Structures: Standard Search (for exact brand protection and precise intent) and Standard Shopping (for granular product control) retain their strategic justification alongside PMax. By 2026, Performance Max campaigns are no longer the non-transparent black box that SEA managers complained about in the early days. Google has made massive technological upgrades and given advertisers tools that allow for fine-grained adjustments. These include campaign-wide negative keywords, optimized search term reports, transparent channel performance reports, deep asset metrics, segmentable reports for asset groups, as well as advanced demographic exclusions and device controls. Google's internal data shows that over one million advertisers now use PMax structures. Despite this technological maturity, a fundamental principle remains: a Performance Max campaign never optimizes itself in terms of your actual business model. The system operates purely opportunistically based on the data provided to it. If an unqualified, faulty contact form is counted as a successful conversion, the artificial intelligence scales exactly those low-quality lead sources. If expensive brand traffic artificially inflates the Return on Ad Spend (ROAS), the algorithm gratefully grabs it without generating real incremental revenue. For demanding SEA managers and marketing decision-makers, this means that optimization today no longer takes place primarily via manual bids, but through strategic data management, placing precise guardrails, and honest performance measurement. Deeply Analyze Budget Distribution and Channel Performance As soon as a Performance Max campaign shows a drop in performance, many market participants tend to immediately modify the target ROAS (tROAS) or target cost-per-conversion (tCPA). In practice, this lever is usually pulled too early and only treats symptoms instead of causes. The first analysis step must absolutely be looking at the budget distribution across the various networks. The dedicated channel performance report reveals which budget shares are flowing into the Search, Shopping, YouTube, Display, Discover, Gmail, Maps channels or to search network partners. Although this report does not allow for direct, manual budget reallocation, it makes dangerous shifts transparent. If, for example, spending in the Display or YouTube network suddenly spikes and at the same time the final lead quality in the customer relationship management (CRM) system drops, the cause is not an incorrect bid level. Rather, the campaign is attracting low-quality clicks through visual placements because the underlying conversion signal is too weak or too easily manipulated. As part of a deeper Performance Max optimization, search terms must be consistently analyzed and prioritized by total cost. Frequently, expensive search queries without any conversion action are much more revealing than historical winners. SEA managers should systematically identify and exclude unsuitable search terms. Typical negatives that should be placed in almost every professional B2B or e-commerce account include terms like: "jobs," "career," "salary," "support," "login," "free," "guide," "PDF," student research, irrelevant competitor names, or purely informational search phrases with no commercial intent. Strategic Campaign Structure by Profitability In many accounts, the structuring of Performance Max campaigns follows purely visual or catalog-based criteria. This is inefficient. A split into separate campaigns is only justified if this split enables targeted operational control – be it through differentiated budgets, specific target bids, differing conversion goals, margin structures, regional focus areas, or strict brand rule sets. Segmentation Criterion E-Commerce Approach Lead Generation Approach Profitability & Margin Splits by high-margin (e.g., private labels) vs. low-margin (retail goods). Focus the budget on products with real return. Differentiation by Customer Lifetime Value (CLV) or order volume (e.g., enterprise deals vs. SMB self-service). Product & Service Dynamics Separation of bestsellers (high-performers), seasonal goods, new arrivals, and so-called zombie SKUs (products without clicks). Differentiation between high-margin core services and purely informational introductory offers (e.g., whitepaper downloads). Database (Custom Labels / CRM) Steering via the Google Merchant Center feed using defined custom labels for inventory and margin classes. Steering via verified offline conversion data (MQL, SQL) instead of pure online form submissions. The exact same economic principle applies to lead generation. Segmentation must be based on sales reality. Never structure your asset groups or campaigns primarily on audience signals. Since Google only interprets these signals as a non-binding recommendation, a purely audience-based campaign separation almost always leads to internal data overlap and inefficient budget allocation. Align Search Themes, Audience Signals, and Customer Match Precisely The introduction of search themes offers an excellent option for sharing contextual knowledge with Google AI. However, search themes should never be confused with classic keyword match types or seen as a complete replacement for structured search campaigns. Their strategic area of application is primarily where the system has too little historical data: during the market launch of completely new product lines, for highly complex B2B niche applications, for targeted promotion of competitor alternatives, or when the landing page offers too little semantic text content due to a minimalist design. Even though Google allows up to 50 search themes per asset group, this limit should never be maxed out randomly if you want precise Performance Max optimization. Best practices suggest using a few, concise themes bundled strictly by search intent. Afterwards, the generated search term reports must be closely monitored to immediately prevent any misdirection of the algorithm. The same applies to audience signals. They do not represent a hard, exclusive target, but rather act as an initial catalyst for machine learning processes. Advertisers should consistently rely on first-party data here. You will achieve the highest signal quality through: Up-to-date customer match lists from your CRM (high-value buyers). Granular website visitors (cart abandoners, returning users). Specific app user data or qualified newsletter subscribers. Isolate Brand Traffic and Secure Incremental Growth It is one of the most common phenomena in SEA practice: a Performance Max campaign delivers outstanding ROAS metrics on paper, but real company growth stagnates. The reason lies in the uncontrolled skimming of existing demand. The system tends to target brand search queries (brand traffic), existing remarketing audiences, and loyal customers who would convert anyway in order to easily meet predefined efficiency targets. Although Google prioritizes identical exact match keywords in regular search campaigns over a parallel PMax campaign, as soon as the search campaign hits a budget limit or is restricted by settings that are too tight, PMax takes over the brand auction. SEA managers must therefore check at regular intervals which search terms are being actively triggered within PMax and whether unwanted cannibalization effects are occurring with existing brand, generic, or competitor campaigns. To drive genuine, incremental revenue, brand exclusions should be implemented directly in the campaign settings. For e-commerce, specialized search-only brand exclusions are also available. This feature suppresses pure text ads for brand terms within PMax, but still allows the algorithm to display visual brand shopping, which is highly profitable in most cases. Optimize Data Quality in the Feed and Final URLs Particularly in retail, Performance Max is often structurally much closer to a classic shopping campaign than an all-encompassing multi-channel campaign. Before making far-reaching bid adjustments, absolute data quality must be ensured in the Google Merchant Center. Optimizing product titles, product types, GTINs, high-resolution imagery, correct sale prices, precise stock status, and custom labels forms the bedrock. Product titles should not simply be copied from internal ERP systems. They must include the attributes that customers are actively searching for. The optimal layout usually follows this logic: Brand + Product Type + Model Number + Material + Specification (e.g., size, color, compatibility). An often overlooked pitfall lies in the uncontrolled activation of final URL expansion. This feature allows Google to replace the destination page with a supposedly more relevant URL on your website and automatically generate matching text assets. With a brilliantly structured, purely sales-oriented website architecture, this delivers excellent results. However, the setup becomes highly inefficient if informative blog posts, support documentation, career pages, or general advice articles unintentionally slip into the ad pool. Such URLs must be consistently blocked using explicit exclusion rules. Link Bidding Strategies to Qualitative Conversion Signals Choosing the right bidding strategy largely determines the success of a campaign. In e-commerce, the "maximize conversion value" strategy combined with a defined target ROAS is the gold standard – assuming revenue values are transmitted to the Google Ads account perfectly and without delay. A target ROAS that is selected too aggressively starves the algorithm of necessary liquidity and chokes campaign volume. A target value that is set too low generates massive revenue but is no longer economically viable at the margin level once all costs are considered. In the B2B segment and for lead generation, the exact definition of the conversion action is even more important than the bidding strategy itself. If you define the simple submission of a contact form as your primary conversion, you force PMax to maximize exactly these quantitative completions. The result is often a flood of spam leads or contacts with no real interest in buying. The solution lies in shifting optimization to qualified, deeper-funnel offline conversions via CRM import. Optimize for: Marketing Qualified Leads (MQL) after successful initial vetting. Sales Qualified Leads (SQL) after direct sales contact. Generated pipeline opportunities or final "closed-won" deals. A seemingly cheap Cost-per-Lead (CPL) that does not lead to measurable sales is not a marketing success; it feeds machine learning with useless training material. Validate Incrementality Using PMax Experiments Because Performance Max is excellent at funneling existing demand channels, evaluation must never occur in the silo of the campaign dashboard. SEA managers must isolate the real added value (incrementality). The integrated Performance Max experiments are ideal for this. Google provides these as scientific A/B tests with which strategic settings, creative directions, or completely new campaign setups can be compared in a statistically clean manner. Specific uplift tests also precisely measure the real additional benefit of PMax in direct comparison to already active search, video, and display campaigns. For a valid implementation in marketing practice, the following basic rules must be observed: No testing during peak seasons: Never run experiments during extreme seasonal fluctuations (e.g., Black Friday or the holiday shopping season). Single-variable principle: Never change the feed, budget, and bidding strategy simultaneously within a test run. Allow sufficient runtime: Do not cancel experiments after just a few days; the algorithm needs an adequate learning and consolidation phase. The ultimate success criterion is never the isolated ROAS of a single campaign, but whether the overall revenue, net profit, and qualified sales pipeline of the entire company increase significantly. The Continued Relevance of Standard Search and Standard Shopping Despite the omnipresence of PMax in 2026, switching your entire advertising account to this campaign type would be a fatal strategic error. Traditional campaign formats retain their fundamental place in a balanced overall strategy. Classic standard search campaigns (Standard Search) are still indispensable for seamless brand defense, targeted and aggressive bidding on competitor keywords, highly regulated advertising claims, and specific B2B search queries with high exactness. Using exact match keywords ensures that the text ad written correlates perfectly with the user's search intent – a level of precision that PMax inherently cannot guarantee. Similarly, Standard Shopping remains an incredibly powerful tool for tactical product control. When it comes to realizing targeted clearance sales, boosting so-called shelf warmers (zombie SKUs) with a specific budget, quickly reducing inventory, or running highly time-limited promotions for exclusive SKUs, Standard Shopping offers the required granular control at the product level. In the most successful ad accounts of 2026, a hybrid account model has been established: PMax serves as a scale-strong foundation for broad market coverage, Search secures high-quality intent, and Standard Shopping is used for surgically precise feed control. The Paradigm Shift: From ROAS to POAS (Profit on Ad Spend) The classic Return on Ad Spend is increasingly reaching its limits in modern e-commerce. It is a pure revenue metric. ROAS suggests success where financial losses may actually be occurring, as it completely ignores real gross profit. A product that generates $200 in revenue at a 20% margin must be evaluated completely differently from a business perspective than a product that generates $200 in revenue at a 60% margin. Purely revenue-based bidding treats both scenarios identically. This is where the concept of Profit on Ad Spend (POAS) comes in. This metric relates the actual profit achieved to the advertising spend invested: POAS = Gross Profit from Ad Investment / Ad Cost To implement profit-based bidding in Performance Max, detailed shopping cart data and exact cost of goods sold (COGS) must be transmitted to Google Ads via the Google Merchant Center. Since PMax is naturally designed to realize the maximum conversion value within budget, the system runs the risk of heavily scaling low-margin bestsellers without this profit context, while neglecting highly profitable products due to a lack of initial search volume. A high ROAS does not protect against declining overall profitability. Conclusion: Set Guardrails and Keep the AI Under Control In 2026, Performance Max stands out as a highly sophisticated, excellently controllable marketing tool. The main task of SEA managers and marketing executives is no longer manually rebuilding every single ad auction. Your primary responsibility lies in defining crystal-clear guardrails. You must define where the algorithm is allowed to learn – and where it is rigorously blocked. Those who intelligently combine data quality, technological controls, and business logic like POAS will transform Performance Max from an unpredictable black box into a highly profitable growth engine. FAQ on Performance Max Campaigns 2026 Should PMax completely replace Standard Search in 2026? No. Performance Max is excellent for unlocking additional reach and incremental placements. However, it by no means replaces dedicated search campaigns where you need absolute control over keywords, exact ad copy, and the protection of your own brand. Are audience signals in PMax equivalent to hard targeting? No. Audience signals are purely guiding aids for Google AI to speed up the learning phase. They do not restrict ad delivery exclusively. To maximize signal quality, you should consistently feed in first-party data such as customer match lists, CRM segments, and deep website interactions. When is it advisable to use PMax experiments? Using them is highly recommended whenever you want to test the incrementality of your campaigns. Experiments show you in black and white whether PMax is generating genuine new revenue or merely claiming conversions that would have come in anyway through organic search or existing search campaigns. Why is ROAS losing importance as a primary metric for PMax? Because ROAS only measures the ratio of revenue to cost. Since PMax operates autonomously, it optimizes for revenue volume. If your product range has varying margin structures, this often leads to unprofitable products being pushed. POAS (Profit on Ad Spend) is the much more honest business metric here. How often should Performance Max optimization take place? A weekly rhythm is recommended for controlling the channel mix, evaluating search terms, adding exclusions, and reviewing landing pages. Comprehensive audits of brand exclusions, analysis of SKU concentration, updating assets, and reconciling with CRM data should be carried out monthly.
Show more posts