What is query fan-out?
When a user asks an AI engine “is a heat pump worth it for a 1970s house in a cold climate”, the engine does not search that sentence once. It fans the question out into sub-queries: heat pump efficiency at low temperatures, installation cost, retrofit requirements for older homes, subsidy programs, comparisons with gas boilers, owner satisfaction. Each sub-query retrieves its own candidate passages; the surviving material is grounded, merged and cited. Google confirmed this fan-out behavior publicly in 2023, and it is the first stage of the four-stage pipeline shared by modern AI search.
Why does fan-out change content strategy?
Classic SEO targeted a keyword; fan-out means the engine targets a topic graph. A page can rank number one for the head query and still be absent from the AI answer, because the answer is assembled from passages that won sub-queries you never covered. Conversely, a site with modest classic rankings can dominate AI answers if it covers the fan-out tree comprehensively. This is exactly what the Semantic Coverage dimension of the open measurement standard captures: what share of the sub-question space does your content answer well enough to be retrieved and cited?
How do you map a fan-out tree?
- Start from real prompts. Collect the questions buyers actually ask AI engines about your category — sales calls, support tickets, community threads.
- Decompose. For each head question, list the sub-questions an answer requires: definitions, costs, comparisons, risks, how-to steps, local or segment variants.
- Observe the engines. AI Mode exposes related searches; Perplexity shows follow-up suggestions. Both are direct windows into the fan-out tree.
- Score coverage. Mark each sub-question: answered on your site, answered by third parties citing you, or unanswered. The last bucket is your content roadmap.
How do you cover the tree without bloating your site?
- One page per intent cluster, with a self-contained answer-first section per sub-question — not one thin page per sub-query.
- Link sections together so crawlers and retrieval systems see the cluster as a coherent whole. See site architecture for AI.
- Use the mapping discipline from semantic coverage mapping to track the tree as a living document.
- Refresh coverage when engines add sub-questions — new regulations, prices or model generations create new branches.
Common pitfalls
The most common mistake is covering only the “money” branch — the purchase-decision sub-queries — while competitors own the definitional and comparison branches that shape the answer’s framing. Second, teams produce near-duplicate pages for every sub-query, which confuses retrieval and dilutes authority; depth beats count. Third, fan-out trees differ by language and market; a tree mapped in English does not transfer to other locales without rework, as covered in multilingual GEO. Finally, remember fan-out is dynamic: re-map quarterly as engines and questions evolve.
Treat the fan-out tree as shared infrastructure across teams. SEO uses it for content planning, product marketing uses it for messaging gaps, sales uses it for objection handling, and support uses it for documentation priorities. When one tree serves four teams, coverage gaps get fixed faster — and the Semantic Coverage score becomes a company metric rather than a marketing curiosity.