Skip to main content
When you ask an AI assistant a question, it often does not search for your exact words. It breaks the question into several searches of its own, reads the results and writes an answer from them. Those searches are called fan-out queries. Trackbase records them for ChatGPT and Gemini, on the AI Visibility page in the Fan-outs tab.

​
Why they matter

The prompt is what your customer asks. The fan-out queries are what the model looks up. The pages that rank for those searches are the pages that end up in the answer. For example, a prompt such as “which accounting software is best for freelancers” can lead to searches like “best accounting software freelancers 2026” and “accounting software reviews comparison”. If your content answers the prompt but not those searches, the model may never find it.

​
What the tab shows

  • A summary: the number of unique queries, how often they occurred and how many prompts have fan-outs.
  • A table with each query, the models that ran it, how often it was seen, when it was last seen and the prompt it belongs to.
  • A filter per prompt, and a switch between all models, ChatGPT and Gemini.
  • A copy button per query.
Fan-out queries are collected from the moment a prompt runs on ChatGPT or Gemini. A new prompt shows its first fan-outs after its first runs.

​
How to use them

  1. Open the fan-outs for a prompt where you are missing.
  2. Note the words the model adds to the question, such as a year, “reviews”, “comparison”, “prices” or a place name.
  3. Check whether a page on your website answers those searches directly.
  4. Add what is missing to an existing page, or write a new one. In Content, the same searches are shown per opportunity under what the models searched for.
Fan-out queries also appear on the Shopping page, for the prompts that return product results.
Last modified on November 12, 2026