What is query fan-out?

Query fan-out is a retrieval technique that AI search engines such as Google AI Mode, AI Overviews and ChatGPT search use to answer a prompt: the model splits the prompt into several related sub-queries, runs those searches, then combines the results into one answer with cited sources.

Also called: Sub-queries, Synthetic queries

Why query fan-out matters for your brand

With query fan-out, the prompt a user types is not the only search that shapes the answer. The engine also runs searches the user never sees, and the pages that rank for those hidden searches are the pages it can cite.

The effect shows up in the data. In a March 2026 study of 863,000 keyword searches, Ahrefs found that only 37.9% of the URLs cited in Google's AI Overviews also appeared in the top 10 results for the query itself, down from about 76% in July 2025. The rest came from positions 11 to 100 (31.2%) or from beyond the top 100 (31.0%). Ahrefs links the shift to fan-out: AI Overviews increasingly cite pages that rank for the sub-queries.

For a brand, this cuts both ways. Ranking first for a head keyword no longer guarantees a citation, and a page that ranks for one narrow sub-question can be cited even if it is invisible for the main query.

How query fan-out works

Google defines query fan-out 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." In practice, the process has four steps:

Interpret the prompt. The model works out what the user is really asking, including implied criteria such as budget, location or use case.
Generate sub-queries. It writes several searches, each covering one angle: definitions, comparisons, prices, reviews, recent news.
Retrieve results. It runs those searches and collects the pages that come back.
Synthesize and cite. It writes one answer from the retrieved pages and links to the ones it relied on.

Google confirms that both AI Overviews and AI Mode "may use a 'query fan-out' technique — issuing multiple related searches across subtopics and data sources — to develop a response." When it launched AI Mode, Google said the technique breaks a question down "into subtopics" and issues "a multitude of queries simultaneously." ChatGPT works in a similar way: OpenAI explains that ChatGPT search "typically rewrites your query into one or more targeted queries" before sending them to its search providers.

Example: one prompt, several searches

Take a prompt a small business owner might type into an AI assistant:

What's the best CRM for a 10-person sales team?

An engine that fans out does not search for that sentence alone. A good answer needs several angles: a shortlist of CRMs that suit small teams, their price per user, head-to-head comparisons, reviews from sales managers and how hard each tool is to set up. Each angle can become a separate search, run in parallel.

Each sub-query returns its own results. A vendor with only a homepage and a pricing page competes on one or two of these searches at most. A vendor that also appears in comparison articles, review sites and setup guides can be pulled in from several of them.

Developers can see real sub-queries. When the Gemini API answers with Google Search, it reports the searches it ran: in Google's documentation, the prompt "Who won the euro 2024?" leads to the search "UEFA Euro 2024 winner".

Query fan-out vs keyword research

Keyword research and query fan-out both start from what people want to know, but they compete on different units.

	Keyword research	Query fan-out
Who writes the query	A person typing into a search engine	The AI model, on the person's behalf
Can you see it	Yes, in search volume tools	Only partly, through APIs and logs
Unit of competition	One keyword, one results page	Many sub-queries behind one prompt
What wins	The best page for that keyword	Pages that answer each sub-question clearly

How to find and target fan-out queries

List the sub-questions behind your key prompts. For each question your buyers ask AI assistants, write down the angles an engine needs to cover: definition, comparison, price, reviews, setup, alternatives.
Check who ranks for each sub-question. Search each one on Google and Bing. The pages that rank there are the pages AI engines can retrieve.
Answer each sub-question clearly somewhere. On your own site, with a focused section or a topic cluster of pages, or on third-party pages such as comparisons, reviews and guides that already rank.
Track the answers, not just your rankings. Follow the prompts your buyers ask and read what each engine replies. Hawkscan shows the full answers from ChatGPT, Microsoft Copilot, Gemini and Google AI Mode, the pages they cite, and how your AI visibility score moves over time.

Common questions

Does Google confirm it uses query fan-out?

Yes. Google's Search Central documentation states that AI Overviews and AI Mode "may use a 'query fan-out' technique" to develop a response, and Google described the technique when it launched AI Mode in May 2025.

Can you see the searches an AI engine runs?

Sometimes. Developers can: the Gemini API reports the searches the model ran with Google Search. OpenAI confirms that ChatGPT search rewrites prompts into one or more targeted queries. Consumer products rarely list the exact sub-queries, so most teams work from the sub-questions they can infer and check.

Should I create one page per sub-query?

No. Google's guide to generative AI features names fan-out queries explicitly: creating separate content for every variation of how people might search, including fan-out queries, primarily to manipulate rankings or AI responses violates its scaled content abuse policy. Cover related sub-questions inside focused, genuinely useful pages instead.