Query fan-out generator

A modern answer engine rarely retrieves against the sentence you typed. It decomposes the question first — into a definition lookup, a comparison, a price check, a “does it work with X” — retrieves for each, and writes one answer over the results. That decomposition is the fan-out, and it is why a page that ranks for your headline question can still be absent from the answer: the retrieval that mattered was for a sub-question your page never addresses.

This page shows a modelled fan-out. It is not a recording of anything. Your prompt is sent, once, to a general-purpose model under a response schema that permits nothing but a list of short queries each tagged with one facet from a fixed set. What comes back is that model’s decomposition. It is not ChatGPT’s query log, it is not Google’s AI Overview trace, and it is not a reconstruction of either. Those are private to the companies that run them.

The facet names are ours, not any engine’s. We hold the model to a closed list — definition, capability, comparison, alternatives, pricing, use case, integration, implementation, evaluation — so that two runs are comparable to each other. An open facet field would let the model invent a new grouping every time, which reads as insight and is noise. Anything it returns outside the list is shown under “Other” and labelled with what it actually said.

There is no score here and no volume estimate. We do not consult search volume data for these sub-queries, so any number claiming how often each is asked would be invented, and the invented part is the part you would act on.

The useful reading is rarely the individual wording. It is the spread. Three things are worth looking for:

• A facet with several sub-queries and no page. If the expansion keeps asking about integrations and you have no page naming the tools you connect to, that is a retrieval you cannot win. Write the page; the sub-query is the brief.
• A comparison you did not choose. Fan-outs surface the competitors a model considers adjacent to you. Being named beside someone is not a ranking — but it does tell you which comparison page a reader is one hop away from.
• A facet the model never reached. Run it two or three times. A facet missing from every run says the prompt itself does not carry that dimension, which is usually a sign the prompt is narrower than the market it stands for.

A sub-query you disagree with is still information. It means the phrasing of your prompt pulled the model somewhere you did not intend, and a real user typing that prompt would be pulled the same way.

Classic search rewarded one page for one query. Answer engines split one question across several retrievals and then cite whichever source answered each part best — which means the page that gets cited is often not the page that best answers the whole question. It is the page that best answers one slice of it, cleanly enough to be lifted.

That has a practical consequence for how you write. A single long page covering nine facets shallowly loses to nine pages that each answer one facet completely, because retrieval happens per sub-query, not per topic. The fan-out is the closest thing to an outline of what those pages should be — and once they exist, the question stops being “do we rank” and becomes which prompts name us and which name someone else.