Not "AI-SEO", but normal text: a study of how Runiti's AI reads websites and what it changes in SEO

Over the past year, SEO has changed significantly. It's not about new ranking factors or another algorithm update, but about how search works now. Users receive answers to their queries directly in the search results, and having a website in the top rankings no longer guarantees clicks and stable organic traffic.

We see these changes not only at the market level but also in the data from our own projects. That’s why we decided to independently investigate how AI-generated answers in search influence clicks, CTR, and user behavior regarding informational queries. In this article, we will share the results and conclusions.

How search has changed: from a link directory to ready-made answers

A few years ago, SEO operated on an extremely simple logic. You publish a page, it rises in the rankings, the user clicks on the link—and gets the information. This model was the foundation of internet marketing and was measured by positions, CTR, and the number of clicks.

In 2025, this logic began to break down. Search engines ceased to be merely navigation tools and increasingly began providing answers directly in the search results. Users no longer need to visit a website to obtain the knowledge they seek. This was driven by the active integration of AI into search. The most noticeable influences on traffic now include:

• AI Overviews in Google;

• search in Yandex with Alice and neural network blocks;

• AI assistants and corporate models like ChatGPT, Perplexity, and others.

In all these scenarios, the website's content can participate in forming responses, but clicking on a link is no longer a necessary step.

What we researched and what we saw

To understand how these changes affect organic traffic, we within the Runity team studied the impact of AI responses in search on informational queries and the dynamics of the organic channel. In our analysis, we observed several consistent effects.

Firstly, the CTR for informational queries decreased by about 10–30%, even while maintaining positions in the search results. Secondly, for queries where an AI response appears, traffic under similar conditions decreased by 1.5–2 times.

It is important to clarify: it is currently impossible to accurately measure the CTR of neural responses. The market does not yet have tools that would allow for correctly linking a website's presence in an AI response to clicks. However, based on a combination of indirect signs and observations, it seems that there are very few clicks from such responses. At the same time, we do not see signs of a decline in ranking or loss of relevance. Pages remain in the same positions but receive fewer clicks.

Key Insight: Inclusion in AI Responses Does Not Equal Traffic

One of the most important conclusions from our research is simple but not obvious: inclusion in AI responses does not guarantee traffic.

Previously, high visibility in search almost automatically meant an increase in clicks. Now this connection has been broken. A site can participate in forming a response, be visible in the results, and at the same time lose organic traffic. This changes the very approach to evaluating SEO effectiveness. Positions and visibility no longer always explain traffic dynamics, and a drop in organic traffic can occur without a decline in ranking. From this point on, SEO ceases to be solely about clicks and becomes a story about the role of the site as a source of knowledge that search engines and AI agents rely on.

How AI Reads Websites — And Why This Changes Content Requirements

When we say that AI reads websites, it is important to clarify: it does so not like a classic search robot.

The search engine indexes pages, matches them with queries, and selects the most relevant ones based on numerous signals — from links to document structure. AI operates on a different logic. Its task is to understand the information on the page and retell it in a convenient and coherent manner. To simplify, AI does not seek the best link; it seeks a clear source of meaning.

If the text is hard to retell, AI will not use it

From our observations, AI performs much worse with texts that:

• lack a clear structure;

• have scattered thoughts that are not connected to each other;

• lack clear definitions and answers to basic questions.

If the text is hard to retell, AI is unlikely to use it. In practice, this means that pages which previously could rank due to volume, keywords, or external factors now turn out to be "unreadable" for AI. It simply cannot extract unambiguous meaning from them.

When a Page Loses Value

We have observed that AI much more readily uses pages where the basic things are clearly and consistently revealed:

• what this product or service is;

• who this solution is for;

• why it is needed;

• why the user should pay attention to it.

If these answers are not formulated directly, AI is forced to make assumptions. In such cases, it either distorts the information or excludes the page from the response.

Which Page Elements Become Critical

From practice and analysis, we have identified several elements that are particularly important from the perspective of AI reading:

• clear hierarchy of headings, where each level logically continues the previous one;

• explicit definitions in the format "X is ..." rather than descriptions;

• structured lists, in which each thought is separated and self-sufficient;

• the presence of an author or expert associated with the material.

It is important to note that this is not about some new "AI-SEO." Essentially, these are requirements for well-written and logically organized text, but now they are critical not only for the user but also for AI.

Why it will not be like before

It is worth mentioning the typical mistakes that we are seeing more frequently.

Firstly, this is generative content without internal logic. Such texts may appear voluminous and formally relevant, but AI struggles to extract meaning from them because there is no clear structure and sequence within.

Secondly, this involves contradictions within a single page. If the theses conflict with one another, AI may piece together a response fragmentarily or with distortions — and as a result, the page is either used incorrectly or not used at all.

In summary, AI's requirements for content sound simpler than they seem: the page must be clear, logical, and unambiguous. This does not mean that one should write for AI. On the contrary — one should write in such a way that the meaning cannot be interpreted ambiguously.

FAQ does not work as it seems

When it became clear that AI actively uses website content to generate responses, one of the first pieces of advice on the market was simple: add an FAQ — and you will make it into neural responses. In practice, it turned out to be more complicated. According to our observations, FAQ does indeed work, but not as a universal SEO block and not by itself.

AI does not read FAQ in its entirety

AI does not perceive FAQ as a separate entity. It does not "take" the block of questions and answers as a whole and does not transmit it to the user. Instead, it works with individual questions that appear as natural user formulations and have unambiguous answers. Experts within the team phrase it this way: AI uses not the FAQ format but the questions within it if they:

• sound like real customer questions;

• do not allow for double interpretations;

• have direct and clear answers.

If the FAQ consists of abstract formulations or duplicates the marketing statements on the page, the AI simply ignores it.

FAQ as a continuation of meaning, not a superstructure

The best-performing FAQ questions logically continue the main text of the page. It is not an SEO tail at the end, but a way to explicitly articulate what the user wants to clarify: limitations, nuances of application, typical doubts, edge cases. In this form, the FAQ becomes not a tool for optimization, but a tool for clarifying meaning — and that is what makes it useful for AI.

A little practice: how to sensibly gather B2B pages today

We will immediately clarify: there is no universal recipe, much depends on the condition of the specific page. But assuming that a B2B page is created from scratch, there is a minimum set of things without which it works poorly for both users and AI. In our experience, such a page must have:

• a clear explanation of what the product or service is;

• a clear answer about who the solution is suitable for and who it is not;

• real customer questions that the team has already encountered in sales or support;

• detailed and unambiguous answers to these questions.

It is important that this set is not about SEO tricks. It is about fixing meaning and removing uncertainty. Such pages are considered a reliable source of information by AI and are ranked higher in search results.

Here’s what definitely should not be done

Against the backdrop of active discussions about AI search, there is a temptation to start writing content specifically tailored to algorithm requirements. Our experience suggests that this is a dead-end path.

It seems that the working logic looks different. We write for the user, but we do this as clearly, consistently, and thoroughly as possible — so that the text can be retold without distortion. If the page is understandable to a person, it is generally understandable to AI as well. Attempts to artificially simplify the language, break down meaning for the sake of format, or generate texts without internal logic often lead to the opposite effect — AI either ignores the page or uses it fragmentarily and with distortions.

And what’s next? Less traffic and the fight for trust

If we look at the perspective of the next one or two years, we expect further declines in organic traffic to websites. The reason is that search engines are increasingly providing answers to informational queries directly in the results — through AI-generated responses — so users simply don't need to click on links. Therefore, websites will compete not only for positions but also for being chosen by AI as a source: taking facts, phrasing, and links from them when generating responses.

The role of the technical side and infrastructure

In this context, the technical side of the website begins to play a more noticeable role. According to our estimates, the following become particularly important:

• page loading speed;

• completeness of indexing — it is important that key content is available and not hidden by scripts.

If AI cannot quickly and fully retrieve information from a page, the likelihood of its use decreases sharply.

Final thoughts

The main conclusion we reached during the study is as follows: SEO is no longer solely about clicks. It is increasingly about the role of the website as a source of knowledge.

This is unusual in terms of familiar metrics, but logical in terms of the evolution of search. There is more content, interfaces are changing, and those websites win that can clearly articulate meaning, do not hide important information, and consistently answer users’ questions.

We are interested in how you see these changes from your perspective: are you experiencing a decline in CTR for informational queries, are you revising your approaches to content and page structure, and what role do you currently assign to the website in the search ecosystem? We would be happy to continue the conversation in the comments.