Why the #1 page can lose the citation: what query fan-out is, what Google says about it and how to plan content for the subqueries.
By , founder and lead strategist at Flowup
Direct answer
Query fan-out is the process in which Google's AI turns one question into several simultaneous, related searches to gather more sources before answering. That is why the page ranked #1 for the original question may not be cited: the answer draws on pages that answer the subqueries. Google says that optimizing for this is still SEO, with content that covers the topic well.
Sources verified on September 30, 2026.
The term has been in Google's vocabulary since at least May 20, 2025. In the AI Mode announcement, Elizabeth Reid, VP and Head of Search, described the technique as breaking the question down into subtopics and issuing many searches at the same time, according to Google's blog post on AI Mode. For Deep Search, the same post speaks of hundreds of searches per question.
In 2026, fan-out moved from jargon to documentation. Google's guide to optimizing for generative AI, updated on July 10, 2026, carries the official definition. And the scale has grown: according to Google's blog, AI Overviews passed 2.5 billion users a month, and AI Mode passed 1 billion. On August 28, 2026, Search Engine Roundtable reported that, for some searches, AI Overviews had started to expand on their own into a full AI Mode response.
For anyone doing SEO, the consequence is counterintuitive: ranking #1 for a question does not guarantee a citation in the answer. This guide explains the mechanism, shows what the citation data says, separates what Google asks for from what it says you can skip, and proposes a method for covering the subqueries without sliding into volume SEO.
Query fan-out is the set of searches the AI model generates from a question to bring in more information before answering. Google's guide defines the term as "a set of concurrent, related queries generated by the model to request more information and fetch additional relevant search results to address the user's query." Put simply, it is a group of related searches that the model itself generates and runs at the same time, to bring back more information and more relevant results for the person's question.
In practice, the question a person asks is not the only search that takes place. The system creates others, about parts of the problem, and uses the results of all of them to assemble the answer. Google does not publish which subqueries it generates in each case; the example below is illustrative.
Illustrative example
Question: "what is the best fleet management system for a mid-sized trucking company?" Likely subqueries: fleet management system comparison, fleet management system pricing, ERP integration, tracking and maintenance features, user reviews, fleet management for mid-sized companies. The final answer brings together passages from pages that answer each of them.
| Criterion | Traditional search | Search with fan-out |
|---|---|---|
| What the system searches for | The query the person typed | The query plus several subqueries generated by the model |
| What the person gets | A ranked list of pages | A written answer, with links to the pages that support it |
| How a page gets in | By being relevant to the typed query | By being useful for some part of the question, even without ranking for the original query |
| Unit of work | The keyword | The topic, with its neighboring questions |
| How to measure | Position, clicks and impressions per query | Impressions in AI features per page and citations in the answers |
Citation studies show that a good share of the pages cited by AI systems are not at the top of the original search, and fan-out is the explanation most often given. The numbers vary considerably between assistants and between dates, which calls for a careful reading.
Our reading
Ranking still carries weight, but it is no longer enough. A page that answers the main question well and ignores the neighboring questions competes for only one part of the answer. Whoever covers the topic in depth, across connected pages, has more points of entry, because each subquery is one more search in which the content can appear.
The difference between ranking and being cited, and the steps an AI system follows to choose sources, are in our guide to SEO and GEO. The entry "query fan-out" is also in Flowup's GEO and AEO glossary.
Google treats optimization for AI answers as part of SEO and says that special files and formats are not needed. The guide to optimizing for generative AI says: "From Google Search's perspective, optimizing for generative AI search is optimizing for the search experience, and thus still SEO." In other words, optimizing for AI search means optimizing for the search experience, so it remains SEO.
The message is that fan-out does not call for a new writing technique. It calls for coverage: the neighboring questions need to be answered somewhere on the site, clearly and with sources, so that the subqueries find those answers.
You cannot see the real subqueries, but you can estimate them with good confidence by cross-checking what search shows, what Search Console records and what customers ask. No single source is enough.
| Likely subquery | What the answer needs | Where to answer it |
|---|---|---|
| Comparison between solutions | Criteria and honest differences | Comparison page or a section of the pillar page |
| Price and formats | Ranges and what drives the price | Pricing page |
| Integration and technical requirements | What needs to be in place and how it works | Documentation or technical page |
| Who it is for and who it is not for | Customer profiles and limits | Solution page, "who it is not for" section |
| Proof and reputation | Case studies, numbers and reviews that can be checked | Case studies and review pages |
With the map in hand, the job is to decide what becomes a section of a pillar page and what deserves a page of its own. The rule of thumb is intent: a subtopic with the same intent as the main question becomes a section; a subtopic with a different intent and a depth of its own becomes a page, linked to the pillar.
This design, with a strong pillar page and satellites that answer the neighboring questions, is what we cover in Organic Dominance Architecture. Turning the subquery map into an editorial plan and into pages is the work of Flowup's Growth Content.
The temptation is to create a page for every subquery. Google advises against that path: the guide says there is no need to chop content up, and Google's spam policy, updated on August 28, 2026, defines scaled content abuse as the generation of many pages for the primary purpose of manipulating search rankings, and not of helping users.
Thin pages, which repeat the same information in different words, do not answer the subqueries any better; they only multiply what the system has to ignore. The cost of that model is laid out in volume SEO. Fan-out rewards useful coverage, not the number of URLs.
Measuring fan-out starts with the generative AI features report in Search Console, which Google made available to all sites on August 31, 2026, according to the post on new controls for website owners. It shows the impressions of pages in the AI features of Search, by page and by country. The same post introduces a control that lets an owner take the site out of the AI features, at the cost of the traffic and the impressions that come from them.
For the Brazilian context of AI Overviews, with data and what to do now, it is worth reading AI Overviews in Brazil (in Portuguese). And for the work of being the chosen source, the guide on how to get cited by AI.
What is not yet known
In short
Query fan-out is the process in which AI turns one question into several simultaneous, related searches to gather more sources before answering. Google's guide to optimizing for generative AI defines it as a set of related queries, generated by the model, to request more information and fetch relevant results for the person's question.
No. Long tail is what the person types: a more specific search with less volume. Fan-out is what the system does afterward: it takes the question, short or long, and generates other searches on its own. The subqueries are not shown to the person, and Google does not publish which ones it generates for each question.
Because the answer is not built only from the result of the original search. With fan-out, the system also searches subtopics and may cite pages that answer them better. In the Ahrefs study of March 2026, only 37.9% of the pages cited in AI Overviews were in the top 10 for the original query.
No. Google itself says there is no need to break content into small pieces for AI to understand it. Most subqueries fit as a section of a well-built pillar page. A page of its own makes sense when the subtopic has a different intent and a depth that does not fit in a section. Mass-producing pages risks falling into scaled content abuse.
You cannot see the real subqueries, because Google does not publish them. You can estimate them: related questions in the results, long queries in Search Console, tests in AI Mode that record the subtopics and the sources, and the questions that sales and customer service hear. Tools that simulate fan-out also help, but they are estimates.
According to Ahrefs, assistants such as ChatGPT, Gemini, Copilot and Perplexity also search several variations of the question and combine the results. In the August 2025 study, with 15,000 long-tail queries, only 8% of the pages ChatGPT cited in its text were in Google's top 10 for the original question.
Through the generative AI features report in Search Console, available to all sites since August 31, 2026, which shows impressions per page in Google's AI features. Pages with impressions there, but outside the top results for the main query of the topic, are a sign that the content gets in through the subqueries.
Sources: Google Search Central, guide to optimizing for generative AI (updated on July 10, 2026) and spam policies (updated on August 28, 2026); Google blog, AI Mode at Google I/O 2025 (May 20, 2025) and new controls for website owners (June 3, 2026, updated on August 31, 2026); Search Engine Roundtable, AI Overviews that open in AI Mode (August 28, 2026); Ahrefs, overlap between AI citations and the top 10 (August 11, 2025), AI Overview citations and rankings (July 21, 2025) and AI Overview citations outside the top 10 (March 2, 2026). Accessed on September 30, 2026.
The AI answer is assembled from several searches, and each one is a chance for your content to get in. Knowing which subtopics your site covers, and which it leaves to competitors, is the starting point. Ranking is not enough. Be the answer.
AI SEO
Find out which subqueries of your topic your site does not cover yet
Flowup's free diagnosis shows how search engines and AI assistants read your site. AI SEO turns the subquery map into a content and architecture plan, with no promise of rankings or citations.
Founder and lead strategist, Flowup Agency
Guto Bertoncini is the founder and lead strategist of Flowup Agency, which he has run since 2011. He is the author of the B.I.N.A. Method, Novo SEO and the Base Informacional Semântica (Semantic Information Base), and leads the agency's SEO for AI, GEO and AEO practice, preparing companies to be found on Google and cited by artificial intelligence platforms. He writes about search and AI on the Flowup blog and on his official website.
