Answer Radar

The Query Fan-Out View

Before an engine answers a buying question it usually searches the web, often several times, with queries the buyer never typed. Those searches decide which sources get read. Answer Radar records them per answer and shows them as the fan-out view, which is the closest thing to a keyword list an answer engine will give you.

Quick Answer

Fan-out is the set of web searches an engine issued while answering a prompt. Answer Radar records them per sample and groups them by prompt, then by segment, with how many samples issued each query and how many times in total. Scope it to one run or a period. Two limits are reported: engines may search without reporting it, and Perplexity reports related topics rather than searches.

  • Grouped by prompt, then by segment: engine, instrument, locale.
  • Evidence mode returns every query with the sample ids that issued it.
  • Treat the list as evidence of what was searched, not the complete set.

What is recorded

For every sample, the searches the engine reported issuing while composing its answer, alongside the answer text and the citations. The fan-out tool, answer_fanout_get, returns them for a project grouped by prompt and then by segment, with a count of how many samples issued each query and how many times in total. Scope it to one run with run_id, or to a period with start and end.

This is the retrieval layer between a buyer's question and the answer they get. One question routinely becomes several searches, and those searches are what select the sources the engine reads and cites.

How to read it

Look for the searches that recur across samples and across prompts: they are the phrasings the engine reaches for in your category, and a page that answers them directly is a page the engine can find. Look for competitor names inside searches, which show the engine framing the question around a rival. Look for searches that name a review site or a comparison format, which show what kind of source the engine expects to cite.

Compare fan-out with citation share. A query the engine searched and a domain it cited usually line up; when a heavily searched phrasing produced no citation of yours, that is a specific page to write or fix, and the content gap view will tell you whether an existing page already covers the requirement.

Two limits, reported rather than hidden

A provider may issue searches it does not report, so a fan-out list is evidence of what was searched and not the complete set. And Perplexity reports related topics it suggests rather than searches it ran, so its segments carry a reason and no queries instead of a plausible-looking list we cannot stand behind.

Both limits appear in the tool's own response. A list that looks complete and is not would be worse than an honest gap, which is the same rule the metric layer applies to zero.

  • Fan-out is per engine; do not pool queries across engines.
  • Counts are samples issuing a query, not people searching it.
  • No search volume is attached, because no engine publishes one.

FAQ

Is fan-out the same as keyword research?

It is the nearest thing an answer engine offers, but it is observed behaviour on your prompts, not a market volume estimate. Use it to phrase pages the way engines search, and use Search Console for what real buyers typed.

Why does Perplexity show no queries?

Perplexity reports related topics it suggests rather than the searches it ran. Answer Radar shows a reason and no list for its segments instead of presenting suggestions as searches.

Can I see which sample issued a query?

Yes. Evidence mode returns every query with the sample ids that issued it, and each sample's answer and citations are in the run view.

Related help pages