Query Fan-out

When someone asks an assistant "what's the best AI visibility platform for SaaS brands?", the engine does not run that sentence as a search. It decomposes it into several machine-written queries, searches each, reads what comes back, and synthesizes an answer.

That decomposition is query fan-out. Geonimo captures it from the answers themselves — these are the searches the engine reported running, not a model's guess at what it might have searched.

Why it matters

Fan-out queries are short, unnatural and specific — closer to old-fashioned keywords than to the conversational prompt that triggered them. They are also where source selection happens.

Two practical consequences:

1. A prompt you lose may be losing on one fan-out query out of six. Fixing that one query is a much smaller job than "improving visibility".
2. Fan-out queries make excellent content briefs. They are literal evidence of what the machine wanted to know and could not find well answered.

Three views

By query. One row per captured fan-out query, with how often it was issued, how many of your prompts triggered it, the sources it surfaced, and when it was last seen. Sort by frequency to find the queries deciding the largest share of your answers.

By prompt. Grouped under each tracked prompt, showing the fan-out set that prompt produced. The view for diagnosing a single losing prompt: you see the searches behind it and which returned competitor pages.

By source. Inverted — which fan-out queries surface a given domain. Use it to understand why one domain dominates your category: usually it wins a cluster of related queries rather than one big one.

Search across queries and prompts to jump to a term you care about.

Coverage varies by engine

Not every surface exposes its searches. ChatGPT reports the sub-queries it ran; other engines expose fewer or none, and the tab shows the per-engine breakdown of what was captured.

An empty fan-out list for an engine is a property of that engine, not evidence that it searched nothing. Read fan-out as a window into the engines that open one, and do not compare fan-out volume across engines as if it were a performance metric.

How to use it

• Write to the fan-out, not the prompt. If "ai visibility software reviews" fires behind five of your prompts, a page answering exactly that beats another general overview.
• Find the unowned clusters. Queries where none of your pages appear are gaps with evidence attached, unlike keyword-tool guesses.
• Watch new queries appear. Fan-out shifts as engines update; a query that starts firing this month is an opportunity before competitors notice it.
• Match structure to intent. Comparison fan-out wants tables and named alternatives. How-to fan-out wants steps and short direct answers. Engines extract passages — give them extractable ones.

Where it sits

Prompt  →  fan-out queries  →  search results  →  sources cited  →  answer

Prompts is what your buyers ask. Sources is what the engine read. Fan-out is the step in between that explains how one became the other.