AI Overview, Perplexity and ChatGPT SEO in 2026: how to get to the sources of AI answers

When they tell me “we have lost SEO in Google,” in 2026 this no longer means something about an algorithm update. This usually means that the AI ​​Overview has eaten up the top part of the search results, and traffic has gone to the sources that this AI summary cites. In this article, I’m not looking at “how AI search engines work” (they already write about this), but specifically: how to rewrite existing content so that Perplexity, ChatGPT and Google AI Overview cite it.

Context: I have been blogging dpustovalov.com since the fall of 2024, in December 2025 I rebuilt the structure of articles for AI Overview, in March 2026 I received the first citations in Perplexity on performance queries. The increase in direct traffic after being included in AI reports is 18-25%, depending on the niche. The numbers in the article are from my own measurements and correspondence with 8 colleagues who also did this in 2025-2026.

The main change of the SEO era: the goal is no longer a click. The goal is to be mentioned in the AI ​​response. A click is a byproduct of a mention.

1. What is AI Overview and why does it break classic SEO?

AI Overview is a block at the top of Google results that collects answers from multiple sources. According to BrightEdge, in 2026 it is shown in 84% of commercial queries in the US and in 60-70% of queries in the RU segment via Yandex Neuro (Yandex analogue).

The problem with classic SEO is simple: if the user receives a response from an AI block, he does not click on the blue links. This is a “zero-click search” - a search that ends without going to the site. According to Gartner, by 2028, zero-click will consume 25-30% of the organic traffic that sites previously received.

But there is good news: AI Overview cites sources via links. If your article is included in the sources, you receive brand awareness, a mention, and some users still click to find out more. The goal is shifting: not “to take first place in the search results”, but “to get into the sources of AI answers.”

2. How exactly do AI search engines choose what to cite?

I spoke with two engineers from large AI startups and read the Perplexity and Anthropic research folders. The logic for selecting sources is approximately the same:

Signal 1 - Structured response at the top of the page. If the first 100 words contain a direct answer to the title question, the article is included in the shortlist. If the answer is spread out over paragraphs or begins with the story “a long time ago it was like this in marketing”, it is usually eliminated.

Signal 2 - Unique empirical data. LLMs don't like retelling what is common knowledge. They need numbers that others don’t have: personal measurements, cases with specific results, audience surveys. The article “10 SEO Tips” with general wording has zero chance. The article “My 10 SEO measurements on a project with 50K visitors/month” is a good chance.

Signal 3 - Schema.org markup. FAQ, HowTo, Article, Speakable - all these types of JSON-LD give the LLM the signal "here is the exact answer to question X in format Y." Without structured data, AI reads text as a raw stream and often fails to extract a specific fragment.

Signal 4 - Citation in other materials. If your article is referenced in Wikipedia, in major media outlets, in research blogs, LLM trusts it more. This is backlinks 2.0 - the old authority signal, but now it is read by AI models, not Google PageRank.

3. How to rewrite an existing article under AI Overview - 5 steps

I rewrote 12 blog articles using this method in the winter of 2025-2026, the average increase in citations was 3-7 per article in the first 30 days. Steps below.

Step 1. Add a TL;DR block in the first 100 words. 40-60 words of a direct answer to the main question of the article. Not “introduction”, not “context”, but literally: “the short answer is X because Y.” AI search engines cite this section.

Step 2: Rewrite each H2 in question form. It was: “Structure of the performance team.” It became: “What is the structure of the performance team in 2026.” Under each H2, the first paragraph should be a direct answer, followed by disclosure. LLMs parse this kind of structure best.

Step 3. Add Schema.org markup - FAQ + HowTo + Speakable. FAQ schema provides an extended snippet with expanding questions in the results. HowTo - provides rich snippets with step-by-step instructions. Speakable - indicates the TL;DR block as key for voice assistants.

Step 4: Include tables and comparisons. A table of 3-5 lines with specific numbers increases the chance of citation by 2-3 times. LLMs love structured data more than dense text. From my own experience: an article with 2 tables is cited approximately 2.5 times more often than the same article without tables.

Step 5. Add unique empirical data. Numbers that only you have: personal measurements, client cases with numbers, surveys. This is the main filter that cuts out 80% of competing content: the majority writes general, and you write specific.

4. AI Visibility Tracking Tools

To understand whether you fall into AI Overview and Perplexity, you need tools. For 2026 the working stack is like this:

A realistic start is a manual check through Perplexity for 20-30 target requests. Record which domains are cited. This gives an understanding of the competitive field in an hour of work, without subscriptions.

5. My own optimization stack

What I do on every new article on this blog to maximize my chance of getting an AI citation:

TL;DR block (40-60 words) with Speakable schema - available on every article starting from December 2025. FAQ schema of 6-8 questions - added to pillar articles. HowTo schema - where there are step-by-step instructions. At least 2 tables with numbers in each article longer than 2000 words. Unique data from my 6 current projects or correspondence with colleagues. Internal links to 3-5 of my articles are for building a topic cluster.

Read more about prompts for content generation, about bypass AI detectors, about SEO in 2026 for performance, about setting up Ya.Metrica and AI team transformation.

6. Conclusion

AI Overview, Perplexity and ChatGPT are not the end of SEO, they are a change of purpose. Previously I fought for a click, now I fight for a mention. Content structure that wins: TL;DR in the first 100 words, H2 as questions with direct answers, structured data, tables with numbers, unique empirical data. Those who rebuild in 2026 will experience growth; those who continue to write “SEO content of 2018” will be gradually displaced.

If you want an analysis of your specific blog/site for readiness for AI Overview, write to Telegram or through form. An hour is free, then depending on the situation.

Related materials: content plan template, CPC benchmarks in Yandex Direct, a complete guide to performance marketing, checking the page for AI results.