From click to AI decision: What Agentic Commerce means for brands

Everything at a glance:

• In 2026, AI agents will handle research, comparison, and in some cases even parts of the checkout process on behalf of users

• According to the internetwarriors GEO study (May 2026): Over 80% of ChatGPT citations do not come from the Google Top 50

• FAQ pages, how-to guides, and comparison tables are the most cited formats in AI systems

• Schema.org markup is becoming a mandatory infrastructure requirement, not just an optional add-on

• AI Overviews reduce the click-through rate of classic search results by up to 67.8% and require a new paid media logic

What Agentic Commerce means for businesses and their visibility

Agentic Commerce describes the shift from a click-driven e-commerce model to a system where AI agents research products, evaluate options, consider constraints, and prepare specific purchase suggestions. In this model, the online shop is no longer just a sales space, but also a data source, a basis for decision-making, and a transaction infrastructure.

From a technical standpoint, this development is accelerated by new protocols and standardized interfaces. In 2026, the Model Context Protocol (MCP), the Agentic Commerce Protocol (ACP), and the Agent Payments Protocol in particular will become more visible, as they are designed to make context, commerce data, and payment approvals more accessible to AI systems.

The separation between discovery and checkout is key here. Shopify describes Agentic Storefronts in a way that products become discoverable in AI channels via the Shopify Catalog, while the final purchase can take place either in the shop or directly in the respective interface, depending on the channel. It is precisely this decoupling that changes the logic of digital commerce: visibility, recommendation, and checkout no longer need to happen on the same interface.

GEO instead of just SEO: What the internetwarriors study shows

The third GEO study by internetwarriors shows that classic SEO visibility and AI visibility only overlap to a limited extent. For the study, 240 prompts from 12 industries in Germany were analyzed; a total of 5,317 URLs were included in the analysis, of which 4,794 were unique URLs.

The numbers mark a turning point. Of the URLs linked in Google AI Mode, only 15.6 percent are found in the Top 10 of organic Google searches. For ChatGPT, this figure is even lower at just 9.2 percent. At the same time, over 70 percent of AI Mode links and over 80 percent of ChatGPT citations lie outside the Google Top 50.

These results do not mean that SEO is becoming irrelevant. Rather, they show that GEO follows its own selection mechanisms. Ranking well organically still offers benefits in terms of authority and domain trust, but it does not guarantee being cited by generative systems.

Why strong domains alone are no longer enough

A particularly revealing result of the study concerns the role of strong domains. 51.3 percent of the citations in Google AI Mode and 33.0 percent of the citations in ChatGPT come from domains represented in the Top 10 of organic search – though often with different subpages than in classic Google search.

This is a crucial difference. Classic SEO often rewards the single best URL for a topic. In contrast, generative systems search a trusted domain for the specific page that answers a query most precisely. It is not the strongest homepage that wins, but the most relevant subpage.

As a result, the focus is shifting from keyword placements to topic coverage, entity clarity, and depth of answers. Businesses must not only be visible, but also interpretable as a reliable source for machines.

Which content AI systems prefer

The internetwarriors study clearly shows which page types are preferred in AI answers. FAQ, help, and how-to pages account for 22.8 percent in Google AI Mode and 26.3 percent in ChatGPT. Blog posts follow at 19.4 percent and 17.5 percent respectively, and comparison tables at 10.5 percent and 12.1 percent respectively.

This breakdown makes sense. FAQ and how-to pages provide compact, clearly structured answers. Blog posts offer the necessary context. Comparison tables are particularly valuable for AI systems because they make products, services, or options directly comparable based on specific features.

Classic product detail pages, on the other hand, play a smaller role than many retailers might expect. In Google AI Mode, only 3.5 percent of citations lead to product detail pages, and 4.7 percent in ChatGPT. This suggests that AI systems often prefer aggregating or explanatory pages over isolated product views.

How search intent changes the choice of sources

Search intent also changes content preferences. For informational prompts, FAQ/how-to content and blog posts dominate. In Google AI Mode, FAQ/how-to pages sit at 30.46 percent and blog posts at 26.39 percent; for ChatGPT, they are at 31.63 percent and 23.53 percent respectively.

With transactional prompts, the pattern shifts significantly. Comparison tables, service pages, and homepages gain weight, while product detail pages grow but still do not become dominant. This suggests that AI systems often structure purchasing decisions through consolidated comparison pages first, before individual products play a larger role.

This is an important insight for merchants: optimizing only product detail pages is not enough. Generative search and shopping environments require an additional layer of content consisting of FAQs, comparisons, advisory content, and clear service pages.

Why structured data is becoming a mandatory infrastructure requirement

With the rise of Agentic Commerce, structured data is turning into a vital infrastructure issue. It helps AI systems reliably interpret prices, availability, product attributes, delivery terms, return policies, and organizational details.

This also changes the role of technical SEO. Product, Offer, FAQ Page, Organization, Local Business, and, depending on the business model, Merchant Return Policy data are becoming more important because they make information machine-readable, comparable, and actionable. The more consistently and clearly this data is maintained, the better systems can evaluate a brand or offer.

In essence, it is about transforming a website from just a readable page into a decision-ready source. Agentic commerce rewards good data structures, not just good design.

Shopify and Shopware: How platforms are reacting

The infrastructure of major platforms already shows where the market is heading. With Agentic Storefronts and the Shopify Catalog, Shopify relies on a model where discovery takes place in AI channels and checkout is handled either in the shop or directly within the interface of the respective system, depending on the channel.

As a result, attribution is becoming highly relevant again. Shopify tracks orders from Agentic Storefronts using channel or referrer attribution. Visibility in AI systems is therefore not just a matter of reach, but can increasingly be measured as a commerce channel.

Shopware is moving in a similar direction in May 2026. The new sales channel type for Agentic Commerce, OpenAI product feeds, JSONL exports, and AI referral tracking show that product feeds, data formats, and performance measurement are becoming standard tools for the next phase of commerce.

How AI Overviews shift paid media logic

The rise of generative search interfaces is also changing the logic of paid visibility. When an AI summary already does the research work, users are less likely to click on classic ads or standard organic results than before.

The key statistic: a click-through rate of 19.70 percent without AI Overview drops to 6.34 percent with AI Overview – a relative decline of around 67.8 percent. This figure is more important as a strategic signal than as an exact universal number. It shows how much generative interfaces can disrupt previous click behavior.

At the same time, a new opportunity arises: when brands are cited within the AI Overview, the click-through rate of their paid ads placed below increases by up to 91 percent. This makes it clear why GEO and Paid Media are no longer separate disciplines.

For Paid Media, this does not mean moving away from the existing model, but rather realigning it. Being present in the answer logic of generative systems, in product feeds, and in subsequent decision paths not only improves organic visibility, but also enhances the impact of paid campaigns.

Why B2B is particularly affected

In the B2B sector, Agentic Commerce is potentially even more profound than in B2C. Procurement processes there are based on specifications, approvals, boundary conditions, compliance requirements, and recurring supply relationships. This is precisely why structured information, comparability, and reliable data are so relevant for AI-supported selection processes.

A B2B agent needs to compare not just products, but also delivery availability, certifications, contract options, minimum order quantities, or service levels. Companies that present this info only in PDFs, unstructured tables, or vague marketing speak make it harder for machines to evaluate them. Providers with clearly structured, robust data will gain a massive advantage.

This is why B2B showcases that Agentic Commerce is not just a UX topic. It is an infrastructure, data, and trust project. Simply editing website text without systematically organizing product and service data will often leave a company invisible to the new procurement logic.

What internetwarriors calls the "AI-AI Bias"

As an analytical working concept at internetwarriors, we refer to a specific pattern as the AI-AI Bias: the tendency of AI systems to systematically prefer providers with highly clear, structured, and fact-rich information because this data is easier to process, compare, and reuse with less uncertainty.

This mental model corrects a common misconception: the most emotional brand message does not automatically win; instead, it is often the source requiring the least interpretation. Especially in B2B markets, where products are complex and differences need explanation, this bias can decide which providers make the shortlist in the first place.

The 95:5 rule in the Agentic Web

The 95:5 rule – originally from B2B marketing research by the LinkedIn B2B Institute and the work of Les Binet and Peter Field – simply states that the vast majority of potential buyers are not actively in target purchase mode at any given time. Brands must therefore build long-term memory structures instead of just reacting to immediate demand.

In the context of Agentic Commerce, this logic can be expanded. A brand must be present not only in human minds, but increasingly in the data spaces, knowledge graphs, and trained preference patterns of systems. If you only start organizing your structure, content, and entities at the moment of a specific purchase request, you are often too late.

That is why brand building in the agentic web should not be seen as the opposite of performance marketing. Rather, it is a prerequisite for a brand to appear as a trustworthy source, a preferred domain, or a logical recommendation.

Governance, trust, and transaction security

Delegating purchase decisions to machines significantly increases the demands on governance, authentication, and transaction security. According to recent industry surveys, 78 percent of financial institutions expect an increase in fraud cases driven by AI shopping agents. This is pushing the development of