Query fan-out is a technique in which a generative search engine does not treat the user's query as a single string but breaks it down into dozens of sub-queries, runs them in parallel, collects documents for each of them, and only then synthesises a single answer from them. A question such as "which CRM system is worth it for a small company" is thus internally broken down into sub-queries about pricing, integrations, localisation, reviews, or data migration. This brings about a fundamental shift for content: a page optimised for a single keyword stops being enough, because your chance of appearing in the answer grows with the number of sub-queries you manage to cover. In practice, this means building topically complete pages with clearly separated sections for individual sub-questions, and supplementing them with concrete figures, comparisons, and limits that the model has nowhere else to draw on.
See also: Passage indexing, Information gain, Share of voice in AI responses.
