When someone types a question into Google's AI Mode or ChatGPT, the system often doesn't run that exact sentence as a search. It may break the question into smaller ones, search for each, and write a single answer from what comes back.
That changes what "ranking for a keyword" means. Your page may be pulled into an answer because it matched one of the smaller searches, which the customer never saw. This guide explains how query fan-out works, using only what the companies document and one named study, all checked on November 13, 2026.
The customer asks one question. The machine asks five. Your site needs an answer to the five.
What is query fan-out?
Query fan-out is a retrieval technique in which an AI search system turns one question into several related searches, runs them, and combines the results into one response. The name is Google's.
Google's guide for site owners, AI features and your website (last updated December 10, 2025), says AI Overviews and AI Mode may use it, and describes it as "issuing multiple related searches across subtopics and data sources."
Google used the term on March 5, 2025, when Robby Stein, VP of Product for Google Search, introduced AI Mode. On May 20, 2025, Elizabeth Reid, Google's head of Search, described the steps: AI Mode breaks your question into subtopics and issues many queries at once on your behalf.
The same post introduced Deep Search, which Google says uses the same technique at a larger scale and "can issue hundreds of searches" to build a cited report.
Which AI search tools use query fan-out?
Google names the technique. OpenAI and Anthropic document similar behavior in their own words: the assistant writes its own searches, often more than one, and may search again after reading the first results.
| Product | What the company says | Source |
|---|---|---|
| Google AI Overviews and AI Mode | May use query fan-out: multiple related searches across subtopics and data sources. | Google Search Central |
| Google Deep Search | The same technique, scaled up. It can issue hundreds of searches for one request. | Google, May 20, 2025 |
| ChatGPT search | Typically rewrites your question into one or more targeted queries, and may send additional, more specific queries after reviewing the first results. | OpenAI Help Center |
| Claude (web search tool) | Claude decides when to search, and the search step can repeat several times in one request. | Anthropic documentation |
OpenAI's page gives its own example. A researcher asks about the latest drugs that target a specific protein for cancer, and ChatGPT sends a search partner a shorter, keyword-style query with the year added. That is the pattern to remember: the search that reaches the web is not the sentence the person typed.
Google's guide for site owners gives no number of searches, no list of the searches behind an answer and no formula for how the subtopics are chosen. Treat any tool that predicts Google's fan-out queries as an estimate.
What does query fan-out look like in practice?
In practice, query fan-out turns a long, conversational question into a handful of short searches, each covering one part of it. The example below shows how that could look for a homeowner's question about a plumber.
One question, five possible searches
The question a customer types:
Who should I call to replace a
water heater in Columbus, and
what will it cost?
Searches an AI assistant might run:
water heater replacement cost
Columbus 2026
best plumbers Columbus Ohio
reviews
tank vs tankless water heater
cost
water heater permit Columbus
Ohio
Northside Plumbing official
site pricing
- Cost, with the year. The assistant wants current prices, so it adds a date the customer never typed.
- Reputation. A search about who is good. Review sites and local listings are built to answer this one.
- A comparison. The customer didn't ask about tankless heaters, but a full answer on cost needs it.
- Local rules. A narrow question that a single clear paragraph on your site can answer.
- The business itself. A check of one company's own website for its facts.
Notice what happened to the plumber's chances. A site with one "Water heater services" page competes for one of those five searches. A site that also states a price range, explains tank versus tankless, and mentions the permit competes for four.
How many searches does one question trigger?
It depends on the product and the question. The documented range runs from zero searches for a question the model can answer from memory to hundreds for a research report.
- Google gives no number for AI Overviews or AI Mode. Its I/O 2025 post says "a multitude of queries," and "hundreds" for Deep Search.
- Anthropic gives a planning figure for developers: single-fact questions typically use 1 to 3 searches, and comparative research across several subjects can use 10 or more.
- ChatGPT has the best public data, from a study by Ahrefs.
The Ahrefs study, written by Louise Linehan and published on October 12, 2026, analyzed 66.6 million responses from the free, logged-out version of ChatGPT in August and September 2026. It found that:
- 14.4% of those responses contained fan-out queries. Most answers came without one.
- The responses that did averaged about 1.6 fan-out queries each.
- The year "2026" appeared in 22.9% of fan-out queries in September.
- The word "official" appeared in 13.1% of English fan-out queries in September, up from 10.1% in August.
- In a separate test of 20,000 U.S. prompts, 84.4% of factual questions about pricing, product features or company details cited at least one official source.
Read those numbers with the study's own cautions. Ahrefs notes that free, logged-out sessions search less than paid or reasoning models, and calls the figures a snapshot of one model over two months. They describe ChatGPT, not Google.
Why does query fan-out matter for your website?
Query fan-out matters because it widens the set of pages an AI answer can draw on. Google says the technique lets it show "a wider and more diverse set of helpful links" than a classic web search would.
Three things follow for a small business site:
- Narrow pages get a chance. A page that answers one specific sub-question well can be picked up even if it would never rank for the broad question.
- Gaps cost more. If no page of yours answers the cost question, another site's page fills that part of the answer.
- Your own site is a source about you. The Ahrefs data suggests ChatGPT often looks for a brand's official site when a question is about that brand's prices or features. Facts you leave off your site get answered from someone else's.
None of this needs new technology. Google's guide says there are no additional technical requirements for AI Overviews or AI Mode, and no special optimizations. The page must be indexed and eligible to appear with a snippet. Our guide to showing up in Google AI Mode covers those basics.
How do you optimize for query fan-out?
You optimize for query fan-out by covering the sub-questions behind your customers' main questions, each with a clear answer on a page that search engines can find. The steps below are our working method. They are not a checklist from Google.
- Start with one real question. Pick a question customers ask before they buy. Use their wording, not a keyword.
- Write out the sub-questions. Ask what someone would need to know to answer it fully: cost, options, timing, local rules, who is trustworthy, what can go wrong.
- Map each one to a page. For every sub-question, name the page on your site that answers it. Mark the ones with no page as gaps.
- Fill the gaps with sections first. Most gaps need two or three paragraphs on an existing page, under a heading that states the question. Our guide to structuring content for AI search shows the answer-first format.
- Put your own facts on your own site. Prices or price ranges, service areas, hours, guarantees. State them in plain text.
- Keep dates honest. Update a page when its facts change and show the date. Don't change a date without changing the content.
- Link the pages together. Google lists internal links among its best practices for AI features. Our internal linking guide has a plan for a small site.
Here is step three for the plumbing example, a coverage map you can copy into a spreadsheet:
| Sub-question | Page that should answer it | Status |
|---|---|---|
| What does replacement cost here? | Water heater installation page, with a price range | Gap: no prices on the site |
| Who is a reliable plumber nearby? | Google Business Profile and reviews page | Covered |
| Tank or tankless? | Blog post comparing the two | Gap: no post |
| Is a permit needed? | A section on the installation page | Gap: not mentioned |
| How long does the job take? | Installation page, "What to expect" section | Covered |
Three gaps, and only one of them needs a new page. The other two are a price range and a paragraph. That is typical: fan-out rewards sites that finish their answers, not sites that publish the most pages.
Don't generate a page for every variation of a query. Google's spam policies define scaled content abuse as generating many pages mainly to manipulate rankings and not to help users, however the pages are made.
Google's own advice points the same way. In a Search Central post dated May 21, 2025, John Mueller wrote that people using AI search ask "longer and more specific questions," and that the content to make is unique and satisfying for them.
Can you see or measure query fan-out?
Google's guide for site owners describes no report of the fan-out searches behind an answer, so you measure the result instead. The guide says appearances in AI Overviews and AI Mode are counted in Search Console's Performance report, within the "Web" search type.
That means AI appearances are mixed in with ordinary results. Three checks still help:
- Look for long queries in Search Console. Open the Performance report and scan the query list for long, specific questions. Pages that collect many of them are answering specific questions well.
- Ask the assistants yourself. Put your customer question to AI Mode, ChatGPT and one other assistant. Note which sub-topics the answer covers and which sites are linked for each.
- Track AI referrals. Our guide to tracking AI traffic in GA4 shows how to separate visits from ChatGPT, Perplexity and others.
The second check is the useful one. When an answer covers a sub-topic and links to a competitor for it, you have found a gap for your coverage map.
The short version: one question becomes several searches, and each search is a separate chance to be the source. List the sub-questions, answer each on a page that can be found, and keep your own facts on your own site.
We checked the details in this guide against Google's, OpenAI's and Anthropic's documentation and the Ahrefs study on November 13, 2026. AI search products change often. Covering sub-questions improves your chances of being cited. It doesn't guarantee a citation, and no one can promise one.
Want a coverage map for your business?
Send us the question your customers ask most. We'll map the sub-questions behind it, show which ones your site answers today, and list the gaps. It's part of our AI Search work.
Ask for a Coverage MapFrequently Asked Questions
Where does the term query fan-out come from?
Google used the term when it introduced AI Mode on March 5, 2025, in a post by Robby Stein, VP of Product for Google Search. Its documentation for site owners now uses the same name for both AI Overviews and AI Mode. Other companies describe similar behavior without using the name.
Does ChatGPT use query fan-out?
OpenAI doesn't use the name, but it describes the behavior. Its help page says ChatGPT search typically rewrites your question into one or more targeted queries, and may send additional, more specific queries after reviewing the first results. SEO researchers, including Ahrefs, call these rewritten searches fan-out queries.
Do I need special markup or files for query fan-out?
No. Google says a page only needs to be indexed and eligible to show with a snippet to appear as a supporting link in AI Overviews or AI Mode, with no additional technical requirements. It also says you don't need new machine readable files, AI text files or markup to appear in these features.
Can I see which fan-out searches led to my page?
Google's documentation describes no way to do it. Google says appearances in AI Overviews and AI Mode are counted in Search Console's Performance report within the Web search type, alongside ordinary results. Its guide for site owners doesn't mention a report listing the related searches an AI answer ran.
Should I create a page for every possible sub-question?
No. Google's spam policies name scaled content abuse: generating many pages mainly to manipulate rankings instead of helping users. Answer the sub-questions your real customers have, and answer most of them as clear sections on pages you already own. Add a new page only when a question deserves a full answer.