NEWFree AI visibility report. Tracking from $99/month

Query fan-out

Query fan-out: the searches behind the AI answer.

Query fan-out is the set of searches an AI model runs before it answers a question. CiteHawk captures those sub-queries from the answers it collects, groups them by prompt and checks the concepts behind them against your own site. It runs on every plan, alongside the weekly sweep of up to 8 AI models.

Run my free report

Free AI visibility report · No credit card · 50 prompts · 8 engines

What is query fan-out?

Query fan-out is how an AI model breaks one question into several web searches before it writes an answer. A prompt such as the best software for a job can become searches for reviews, comparisons and pricing. The searches decide which pages the model reads, so they decide which brands it can name.

app.citehawk.com · Query fan-out
The Query fan-out panel in CiteHawk for one prompt: each sub-query the AI models issued, the model that issued it and how many answers carried it, with a note on what each model exposes
CiteHawk's demo workspace, captured 22 September 2026
app.citehawk.com · One answer
One stored AI answer in CiteHawk with its Query fan-out section: the three searches the model issued while answering, above the pages it cited

Captured, not simulated

Query fan-out examples from real answers

Every sub-query on the page is one an AI model issued for your prompt, read from the answer CiteHawk stored. Nothing is generated to fill the gap, so a model that ran no search shows none.

  • The sub-queries behind each prompt, with the model that issued each one
  • How many answers in the run carried each sub-query
  • The queries on every stored answer, next to its citations
  • Concepts and entities named across the searches, each tied to a sub-query or answer

What CiteHawk captures

The searches, the concepts and where you stand

CiteHawk reads fan-out from the AI models that expose it: ChatGPT, Claude, Gemini, and Grok (with the Grok add-on). Perplexity shows the pages it consulted rather than its queries, and DeepSeek, Copilot, Google AI Overviews, and Google AI Mode do not expose their searches at all.

Sub-queries, per prompt

The searches behind each prompt, for up to 60 prompts a run, with the prompts where you are not named taken first.

Concepts with a coverage check

Each concept is checked for the exact term on the site pages CiteHawk has stored.

Sources consulted

The pages the models read, so you can see where the answer came from.

Frequently asked questions

Query fan-out is the set of web searches an AI model runs to answer one question. The model splits the question into narrower searches, reads the pages those searches return and writes its answer from them. Which searches it runs shapes which brands it can name.

Scout leaning over an unrolled map, pinning places

See what AI models searched for before they answered.

Start free. Get your AI visibility report, no credit card required.

See pricing

Free AI visibility report · No credit card · 50 prompts · 8 engines