SEO: Where Artificial Intelligence Plays a Role in Search Engines Now and in the Future

We've already demonstrated in our first part of the series that artificial intelligence is one of the important technologies of the future and has long since become part of everyday life. But how does it look in individual online marketing disciplines? Where are search engines already using AI, and what does that mean for the future of search engine optimization? An overview and outlook.

Google's Development: DeepMind and RankBrain

Ever since DeepMind became part of the Google empire, it's evident how much the search engine giant hopes to gain from its investment in artificial intelligence. According to a report by CBInsights, there is also a significant increase in AI-relevant patents registered by Google, which may be due to the company abandoning its previous reservations about patent filings and entering the competition with major corporations (Microsoft, Google, Facebook, Apple). This is a development that could be viewed critically as it disadvantages smaller companies that cannot compete in the patent race, and it significantly hampers scientific research.
In early 2015, Google announced that it was interpreting a portion of search queries using artificial intelligence. At that time, this involved queries never before asked (long-tail keywords) with the aim of providing relevant search results for these as well. Since last year, the system known as RankBrain has not only influenced these roughly 15% of new queries but all search queries. Since it became known the substantial influence of RankBrain as the third most important ranking factor in the algorithm, the importance of artificial intelligence and self-learning systems in the SEO field has become unmistakably clear.
Incidentally, Microsoft with RankNet shows similar efforts. However, it is not definitively answerable which system delivers better results and understands queries more accurately, determining the intent behind them.

Quality Assessment with AI

A look at Google's development is worthwhile: Even with the Hummingbird Update, the search engine attempted to better understand human language and correctly interpret the intent behind search queries to answer them with matching results. Instead of purely matching strings, as was the case many years ago (when it made a significant difference, for example, whether one searched in singular or plural), Google became increasingly capable, through semantic analyses, of understanding the meaning of terms and delivering results using synonyms of the search query within the content.
RankBrain and artificial intelligence have taken the next step in the interpretation of human language (including spoken language, which is interesting for voice searches or colloquial search queries). Instead of human-created vast databases of related terms, synonyms, and relationships between words, the system is intended to learn independently from examples to understand language and recognize intentions and relationships behind it. This way, patterns between complex search queries are expected to be recognized, which previously appeared unrelated based on standard evaluation methods.
With progress in this area, the search engine's ability to assess content quality improves. This could involve linguistic quality (journalistically demanding texts vs. assembly line written "SEO texts") on one hand, and on the other, content relevance to the target group. This means increasing demands on content production, which faces significant challenges considering the content saturation observed in many industries for several years now.
Simple keyword analyses via the Google Keyword Planner without considering the target group and the intention behind them have long been ineffective strategies. Customer service, marketing professionals, and the social media department must communicate and leverage synergies: What is currently trending in the industry? What topics concern the target group? What problems repeatedly arise with the product, and what purchase reservations exist?
A prime example of thinking outside the box is the evaluation of actual traffic via long-tail keywords, as enabled, for instance, by Google Search Console. Looking beyond the “money keywords” and major traffic drivers is worthwhile: What intentions lie behind awkwardly formulated queries, which Google increasingly strives to interpret and match with precise results? What exactly does the target group need to remain on the page and send accordingly positive user signals—and perhaps even convert in the end?
The more Google learns to understand data and directly reflects it in search results (via markups or the Knowledge Graph), the more it competes. It answers questions itself without needing the user to click a result. Here too, the question arises: What can, what must be offered to visitors to motivate them to visit and stay on the page? A deeper analysis and comprehensive understanding of the target group are prerequisites for success here.

Analyzing User Behavior as a Ranking Factor

It is expected that user behavior will play a greater role than ever in quality assessments. High-quality content and many backlinks alone are not definitive proof that a page actually meets user intent for certain search queries. Larry Kim speculates that machine learning might serve as a supervising entity, verifying alongside traditional ranking factors whether the search result indeed generates clicks and user interaction. Pages lacking or with insufficient user interaction (e.g., brief dwell time) consistently decrease in rankings—he convincingly illustrates this with an example. How Google can achieve this: Through machine learning, continuously tracking and monitoring signals for a variety of long-tail queries. The result is conclusions about the quality and relevance of the respective page and a correspondingly positive or negative ranking development.
Given Google's communicated focus on good "user experience" as a major optimization strategy, this assumption is plausible. What follows from this? The trend continues, and the various disciplines increasingly intertwine: technical on-page optimization, editorial work, and usability have been inseparable for a long time and need to complement each other. The user-centered strategy is not a new idea, but it will continue to gain importance.
Web analysis plays a significant role here: Correct and meaningful implementation helps identify and improve poorly performing pages. Artificial intelligence will undoubtedly contribute greatly in this context by evaluating large amounts of data and deriving actions from it. However, it cannot replace direct contact with the target group and insight into the individual as an individual.
Another suspicion is close: User behavior analyzed through artificial intelligence may not only lead to a global assessment of a page but also influence the already advancing personalization of search results. Already, it is impossible to determine one definitive ranking for a specific query and URL because the position depends on many personal factors of the searcher (e.g., search history) and the device used. It is presumed that employing artificial intelligence will further refine this process.

Video SEO: Interpreting Video Content

In the field of image and video recognition, that is, the analysis of visual media, AI is likely to find a crucial application area. Since the release of Google's project Youtube-8M in September 2016, it's apparent that behind the scenes, efforts have long been underway to machine-identify and categorize the contents of images and videos better.
Similar to the ImageNet database with millions of classified images, Youtube-8M offers 8 million videos, captured using 4,800 entities from the Knowledge Graph. This immense publicly available data set can serve as the basis for machine learning in video, essentially a training set for a system fed with this data.
If the technology for simplifying and evaluating video data continues to improve, it introduces entirely new criteria for video optimization. Previously, descriptions in text form or even transcriptions and manual category assignments were necessary; in the future, the video itself might be interpreted based on its content and ranked accordingly.

Conclusion and Outlook: What Does the Future of SEO Look Like?

Considering the rapid developments in artificial intelligence and the outlook towards a self-learning algorithm evolving by its own rules, the "Black Box" Google becomes even more elusive. Already, there's no longer "the" algorithm. Factors or their weighting differ significantly among various domains and industries. If in the future, the algorithm is even more heavily determined by artificial intelligence, it will be even harder to define ranking factors. This development is assured per Google's statements on the highest priority of AI: from "mobile first" to "AI-first", as Google CEO Sundar Pichai is quoted. The goal is a personalized Google for every user. After two years of "mobile first" with a dedicated mobile index, we can imagine the significant new developments focusing on AI might bring.
Will SEO become obsolete now? By no means, but the field is continually expanding. Tomorrow's search engine optimizer will confront Google's AI algorithm with a tool of their own that aggregates users' behavioral data on the internet and dynamically generates content based on that data, tailored precisely to individual user requirements. The SEO of the future is thus battling Google with its own weapons—AI-based behavior analyses to optimize content!