The question that generative search is trying to answer is not what pages are most relevant to a user’s query. Search has long been able to find relevant pages. The harder question, the one that defines the era of AI search, is: which source, among all of the relevant options, is worth trusting?
Modern search is increasingly built around that question — and at Yandex it’s central to how Alice AI decides what to pull into an answer. In this article, we explain how AI search engines evaluate trust and authority, and how quality patterns shape the content surfaced in generative answers.
Trust Is Inferred Through Patterns
Trust is not something that a brand can easily signal. No single tag or meta field signals credibility to AI search systems. Instead, trust is more likely to be inferred from patterns of quality across the full body of a site’s content, across external mentions, and even across the coherence of how a brand is represented beyond its own domain.
An AI-powered search engine evaluating whether to cite a source is implicitly asking the question “does this material help give the user the best and most credible answer?” The answer lies less in what a source claims about itself and more in everything an AI system can observe about it.
Expertise and authority are treated less as attributes that can be directly measured and more as patterns that emerge from quality. A page that is genuinely useful, original, and expert looks different from one that is only trying to approximate those qualities. Search is no longer just about finding something relevant; it is about finding the best content among many relevant options.
What Patterns Search Actually Looks For
Yandex’s EPOS framework is useful for evaluating the quality patterns that other AI-powered search engines are likely to look for. Expertise, Practicality, Originality, and Substance offer a practical way to judge whether a source genuinely helps with a user’s task — and to tell that apart from content that’s just trying to look high-quality.
Expertise
Expertise is about content with specific details that are only possible from first-hand knowledge of a given topic. Generic material can be assembled from surface-level summaries, but it likely reads differently to a retrieval system than content produced by someone with direct experience in the same domain. What search is getting increasingly good at is distinguishing between firsthand and borrowed knowledge.
Practicality
Practicality is about content that resolves user tasks. There is a measurable difference between a page that merely matches a query and one that resolves what the user is trying to do. Brands that want to be seen as trustworthy by AI-powered search engines need to focus on a specific set of users and publish content that solves their problems with a specific value proposition.
Originality
Originality is about whether content adds anything that other sources don’t. When multiple sources cover the same topic in the exact same way, AI systems have no reason to prefer one over another. Interchangeable sources add less than sources with a genuine point of view, original data, or analysis that the system can’t get elsewhere. To build a comprehensive synthesized answer, AI-powered search engines need unique material that they can actually use.
Substance
Substance is about whether content has depth and factual density. A page that resolves a user’s task will have different levels of specificity and different degrees of engagement with the complexity of the subject than a page that merely covers a topic related to the user’s task. Pages that perform well in traditional organic search but do not appear in generative answers may lack the factual depth that AI systems need to extract.
How the Query Expands
Generative search often handles the space around a query rather than directly addressing only the exact query the user types. Generative systems like Yandex’s Alice AI may expand a query into a set of related questions before selecting sources rather than mapping each query to a single ranked result. The best, most useful, and most credible answer is then synthesized from multiple strong and relevant sources.
The Importance of Full Information Footprints
Understanding trust as a pattern inferred from the full information environment has a direct organizational implication. A company’s site is one data point among many, and users build their understanding of a brand across social platforms, review sites, niche publications, and independent bloggers.
A brand well represented in high-quality third-party content gives an AI-powered search engine more to work with than a brand whose authoritative presence is confined to its own pages. The importance of comprehensive information footprints also means that modern search optimization cannot rely solely on SEO specialists. It relies on product teams, editorial teams, marketing teams, PR, UX, and everyone else who contributes to whether a source earns trust.
Critical Insights
Generative search is a new interface for an old problem: what content is trustworthy and worth surfacing among everything available? The source that best solves the user’s task remains the answer to that question, even if the technology has changed. AI-powered search engines infer trust from patterns of quality across the full body of a site’s content as well as all relevant and quality off-site content.
Author Bio:
Mikhail Slivinskiy is Search Ambassador at Yandex with over 15 years of experience in search technology and SEO. At Yandex, he has worked across product development, webmaster tools, and publisher engagement, including leading Yandex Webmaster from 2017 to 2024. He now focuses on how AI-driven search is evolving and how businesses can maintain visibility through authoritative content.
Read more: How to Build a Thriving Online Business by Choosing the Right Digital Model
