What is query fan-out?

Query fan-out is the technique where an engine breaks the user’s question into several related sub-searches, issued at the same time across subtopics and different data sources, and then brings the results together into a single answer. The term is Google’s own, used to describe how AI Mode works: it "issues multiple related searches concurrently across subtopics and multiple data sources".

Why this changes how you write a page

If one broad question becomes twelve sub-questions, and each sub-question retrieves its own sources, then the unit of competition stops being the page and becomes the answer to each sub-question. A page that answers a broad topic reasonably loses to a page whose blocks each answer one specific sub-question in a self-contained way.

That is the technical justification for three rules that look like style choices: H2 and H3 headings phrased as questions, a complete answer in the first paragraph of each block, and an FAQ with parity between what is visible and what is marked up in JSON-LD. It is not editorial preference; it is alignment with how the engine searches.

How to use fan-out in production

Before structuring a piece, list the eight to twelve sub-questions an engine would likely generate from the parent question. Each one needs its own heading and an answer that makes sense read on its own, outside the context of the rest of the page, because that is how it will be retrieved.

The acceptance test is simple: hand the rendered text to a model and ask it to answer each sub-question using only that content. If the model cannot extract a self-contained answer, the page is not ready, however cleanly it passes a technical validator.

Related terms

Fan-out explains why latent intent matters and why retrieval picks sources block by block, not site by site. See also how ChatGPT chooses the sources it cites.