🔍 SEO / GEO

Query Fan-Out

fan-out queries · query decomposition · AI search sub-queries · fan-out technique

Definition

Query fan-out is an AI search technique in which the original query is broken into a set of sub-queries by subtopic, a search is run for each, and the model builds the answer from the combined results.

How query fan-out works

Classic search answers a query with a single list of results. AI search with query fan-out works in four steps:

The model parses the query and identifies subtopics and implicit questions.
It writes a set of sub-queries for them.
It runs them in parallel against the web index and other data sources.
It combines the results and synthesises an answer with links to sources.

Google described the technique on 5 March 2025, when it launched AI Mode as a Labs experiment: the mode issues multiple related searches concurrently across subtopics and multiple data sources. On 20 May 2025, at Google I/O, the company explained that this lets AI Mode go deeper into the web than a traditional search, and that Deep Search can issue hundreds of searches for one report. The documentation for site owners says fan-out may be used by both AI Overviews and AI Mode. Technically it is an extension of the retrieval stage of RAG.

Example of a query breakdown

Query: “best robot vacuum for a flat with a cat”.

Sub-query	What it checks	Which page answers it
robot vacuum for pet hair	the key specification	category with a filter, buying guide
robot vacuum with self-emptying station	handling hair in the bin	product pages
obstacle avoidance: toys, cables, litter tray	navigation	review, model comparison
robot vacuum noise level	how the pet reacts	specifications on the product page
best robot vacuums rated, reviews	social proof	reviews, roundups
robot vacuum under $400	budget	category sorted by price

The example is illustrative: Google does not publish the real set of sub-queries.

What it means for GEO

Topical completeness beats exact match. Each sub-query is a separate search intent, and the answer draws on sources across all the subtopics.
Build the keyword map around subtopics. Group your semantic core by the questions a shopper resolves on the way to a choice, not only by high-volume keywords.
Sub-queries look like the long tail. Narrow phrasings such as “robot vacuum for pet hair with mopping” are the long tail: they bring little traffic, but they can become a source for the answer.

Important: Google states plainly that no special optimisation is needed to appear in AI Overviews and AI Mode: the page must be indexed and eligible to be shown with a snippet. Fan-out changes not the technical requirements but which content has a chance to be cited.

Practice: a checklist for a store

For each key category, list 8–15 subtopics a shopper checks before buying: specifications, compatibility, care, price, delivery, returns.
Check which of them you already answer: a guide, an FAQ, a comparison table.
Add to product pages the specifications that show up in sub-queries: noise level, dimensions, compatibility, what is in the box.
Run your key queries in AI Mode or AI Overviews and see which sources are cited for each subtopic — that is where the gaps in your content are.

Frequently asked questions

Who uses query fan-out?

Google publicly describes the technique under this name, for AI Mode and AI Overviews. Other AI assistants with web search also run searches while composing an answer, but they do not disclose their mechanics in detail, so equating them with Google's fan-out is only an approximation.

Can I find out which sub-queries my query was split into?

There is no reliable way to see the full set of sub-queries. The sources in the answer help approximate the picture: the cited pages show which subtopics the model considered important. The other route is to collect the subtopics yourself from search suggestions, the People also ask box, questions in reviews and customer support tickets.

What does query fan-out change for e-commerce SEO?

The answer is assembled from sources across different subtopics, so one page built for one keyword covers only part of the picture. The chance of being included grows for a site that answers comparisons, specifications, compatibility, reviews and purchase terms. For a store that means buying guides, comparison tables, FAQs on category pages and detailed product pages.

Is query fan-out the same as RAG?

No. RAG is an architecture in which the model answers from retrieved documents. Query fan-out is a way to organise the retrieval stage of such a system: instead of one query to the index, many are run. Fan-out widens the search, while the principle of generating an answer from retrieved documents stays the same.