Definition
What is query fan-out?
Query fan-out is the technique by which an engine rephrases the question a person asks, breaks it into subtopics, then runs several distinct searches to gather what it needs to build its answer. Google documents it for AI Overviews and AI Mode. A single question from the user therefore triggers several queries against the index.
How the engine fans out a question
A person asks a question. The engine does not only search those words: it rephrases them, spots the sub-questions a good answer must cover, and runs a search for each. It then assembles its answer from the pages found in those searches, and cites some of them.
Google does not display the searches it runs behind the scenes anywhere. Microsoft, on the other hand, publishes Copilot's grounding queries in Bing Webmaster Tools: it is the only place where you can see an engine's real rephrasings.
What it changes for keywords
The classic grid files each keyword under one intent: learn, go somewhere, compare or buy. It still describes the buyer's state of mind. It no longer describes what the engine runs: a person asks one question, the engine asks several.
The unit of production changes with it. A page that covers a real question and its sub-questions is worth more than several pages competing over variants of the same keyword and depriving each other of rankings.
Fictional example: someone asks "which inventory management software should a 20-employee hardware store choose". The engine may separately search for software available in French, its pricing, its link with the point-of-sale system and user reviews. A page that only covers "inventory management software" answers one of those searches. A page that also covers pricing, point of sale and reviews can be picked for several.
Falia draws a production rule from it: one page per real, distinct question, not one page per keyword variant. The questions come from your salespeople and your customer service team, who hear them before and after the sale. A keyword tool gives volumes. It does not tell you whether a query comes before a purchase.
Long questions trigger a generated summary more often: the Pew Research Center measured 53% from ten words up, compared with 8% for one- or two-word searches. They keep their value for presence. A budget funded on a promise of traffic can be disappointed there.
Not to be confused with
- Grounding query
- Query fan-out is the process: breaking one question into several searches. Grounding queries are the searches that come out of it, as Microsoft publishes them in Bing Webmaster Tools.
- Search intent
- Search intent describes what the person wants (learn, compare, buy). Fan-out describes what the engine runs to answer them: several searches for a single question.
Related concepts
- Grounding query
- AI Overview
- Optimizing for generated answers
- Answer-first structure
- SEO writing
- Batch publishing
Further reading
- Search intent: why the grid of four intents is no longer enough
- GEO: what can you measure on Google, Bing and Perplexity?
- Structuring content for ChatGPT and Google: the answer-first approach
- Publishing 100 pages at once on a new site: what it really costs
- SEO writing in the age of AI: what changes and what does not
Related services
Frequently asked questions
Should I write one page for each sub-question?
No. The sub-questions of a single question belong on the same page, each in its own section. Two pages that answer the same question deprive each other of rankings.
Where can I see the searches an engine runs for me?
In the AI performance report in Bing Webmaster Tools, which publishes Copilot's grounding queries. Google does not publish its own: its report gives impressions, without queries.
