What is query fan-out?
Query fan-out is the way an AI search system turns one question into many searches. Instead of running the exact words a person typed, the system works out the subtopics behind the question, issues a separate search for each one at the same time, and writes a single answer from everything those searches return.
Google introduced the term when it launched AI Mode, describing the technique as "breaking down your question into subtopics and issuing a multitude of queries simultaneously on your behalf" [1]. Google's VP of Product for Search, Robby Stein, gave a plain example: asked about things to do in Nashville with a group, the system "may think of a bunch of questions like great restaurants, great bars, things to do if you have kids, and it'll start Googling basically" [3].
So a single question such as:
things to do in nashville with a groupcan become searches the person never typed:
best group-friendly restaurants nashville
nashville bars for large groups
things to do in nashville with kidsThe answer then cites pages found through any of those searches, not only pages that rank for the original wording.
How does query fan-out work in Google AI Mode?
Query fan-out in Google AI Mode uses a language model to plan the sub-queries and ordinary Google Search to run them. Stein described Search being used as a back-end tool: the model generates the related questions, Search executes them in parallel, and the model combines the results into one response with links [3].
Google's documentation for site owners says both AI Overviews and AI Mode "may use a 'query fan-out' technique," issuing multiple related searches across subtopics and data sources [2]. Google adds that while a response is being generated, its models identify more supporting pages, which is why AI answers often show a wider set of links than a classic results page [2].
Three practical consequences follow from that description:
- Sub-queries are real searches. The pages an AI answer can cite are pages Google can already retrieve for some search. Google says there are no extra technical requirements to appear: a page must be indexed and eligible to show with a snippet [2].
- The cited page may not rank for the original query. It only has to be retrievable for one of the sub-queries.
- Answers draw on passages. The model reads chunks of many result sets and stitches them together, so a single well-written section can earn a citation even when the rest of the page is about something else [6].
Is query fan-out used in AI Overviews or only AI Mode?
Query fan-out is used in AI Mode and Deep Search for certain, and Google says AI Overviews "may" use it too [2]. Google has not said how often AI Overviews fan out, or for which kinds of query. The difference matters for measurement: an AI Overview that does not fan out will cite pages from the main results for the query, while one that does can cite pages from other searches.
Search Insider's own data fits the first pattern for short commercial queries. In its stop words study, AI Overview citations moved with the organic results rather than independently, which is what you would expect if most of those AI Overviews leaned on the main query's results.
When did Google introduce query fan-out?
Google first named query fan-out publicly on May 20, 2025, when it announced AI Mode for everyone in the U.S. [1]. Since then Google has added detail in interviews and its Search Central documentation, and the idea has been linked to an earlier Google patent.
December 2024: the Thematic Search patent
A Google patent titled Thematic Search (US12158907) describes generating sub-queries from themes found in the top results, grouping results by theme, and summarizing each group with a language model [4]. Search Engine Journal connected it to AI Overviews and AI Mode in May 2025 [4]. A patent shows what Google has designed, not what it runs, so treat it as background rather than a specification.
May 2025: AI Mode and Deep Search
At the AI Mode launch, Google said Deep Search "uses the same query fan-out technique but taken to the next level," issuing "hundreds of searches" to build a fully cited report [1].
July 2025: Google explains the mechanism
In an interview reported on July 30, 2025, Robby Stein described the model planning related questions and using Search as a tool to run them simultaneously, with Deep Search sometimes running for several minutes [3].
December 2025: written into Search Central
Google's guidance for site owners on AI features, in the version last updated December 10, 2025, states that both AI Overviews and AI Mode may use query fan-out, and that no special files, markup or optimization are needed to appear in them [2].
How many sub-queries does query fan-out create?
Google has not published a typical number of sub-queries for query fan-out. Its only stated figures are "a multitude of queries" for AI Mode and "hundreds of searches" for Deep Search [1].
Some guides state that AI Mode issues a fixed range of sub-queries per question and attribute the range to Google's May 2025 announcement. The announcement contains no such number [1]. Any specific count comes from outside tests that try to infer sub-queries from what gets cited, and those tests cannot see Google's internal queries directly.
What is the difference between query fan-out and query rewriting?
Query rewriting changes one query into a better version of itself, while query fan-out creates several different queries and runs all of them. Search engines have long corrected spelling, added synonyms and reinterpreted wording before retrieving results. That still produces one result set. Fan-out produces many result sets on related subtopics, then merges them.
| Query rewriting | Query fan-out | |
|---|---|---|
| Queries run | One (modified) | Many, in parallel |
| Covers subtopics the user did not type | Rarely | Yes, by design |
| Output | A ranked list | A written answer citing several result sets |
| Where Google uses it | Classic Search | AI Mode, Deep Search, some AI Overviews |
Fan-out queries are one kind of synthetic query: a query generated by the system rather than typed by a person.
Can you see the queries a fan-out generates?
No. Google does not show the sub-queries in AI Mode or AI Overviews, and Search Console does not report them. AI feature traffic is counted inside the ordinary Web search type in the Performance report [2], so there is no way to see which sub-query surfaced your page.
The closest public view comes from Google's Gemini API. When a developer turns on Grounding with Google Search, the model decides whether to search, "generates one or multiple search queries and executes them," and returns a field called webSearchQueries listing the queries it ran [5]. Running a question through the API shows the kind of sub-queries a Gemini model writes for it. This is a proxy, not a window into AI Mode: the API model and settings are not the same as Google Search's production system, so the output shows the shape of fan-out, not the exact queries.
How do you optimize content for query fan-out?
To be cited through query fan-out, a page needs to rank for, or be retrievable for, the sub-questions behind a topic, not only its head term. Google says no special optimization is required [2], which in practice means ordinary SEO applied to the follow-up questions:
- List the sub-questions. For each target query, write down the questions a knowledgeable friend would ask next: cost, comparisons, who it suits, alternatives, local options.
- Answer each in its own section. Use a question heading and answer it in the first sentence, so the passage stands alone when lifted out.
- Check that the page ranks for the sub-queries. If a section is the best answer on the web but the page does not appear for that sub-question, fan-out will not find it.
- Keep facts specific and current. Prices, dates and named options give a model something concrete to cite.
What should you do about query fan-out?
Treat query fan-out as a reason to cover a topic completely, not as a new channel that needs new tools. The pages AI answers cite are found by ordinary Google searches, so ranking for the questions around a topic is still the lever. Be wary of anyone selling a precise count of sub-queries or a tool that claims to show Google's real fan-out queries: Google publishes neither [1][2].