2026-09-15 — GEO · ChatGPT · Fundamentals

Query fan-out: the searches ChatGPT runs before it answers

When ChatGPT searches the web for an answer, it rarely searches the prompt as typed. It runs several rewritten searches, a fan-out, and its cited sources come from what those searches return. Across a set of prompts, the fan-out shows which queries feed the answers in your category.

Request it

{
  "prompt": "What is the best CRM for a small agency?",
  "country": "US",
  "include": { "searchQueries": true }
}

Send it to POST /v1/monitor/chatgpt or as a CHATGPT task. include.searchQueries shares an add-on with rawResponse, ads and shopping: enabling any of them adds the same cost once (pricing).

The fan-out comes back as result.searchQueries, an array of strings. Two details of the contract matter:

• It can be empty. ChatGPT serves more than one interface. "legacy": true requests the legacy desktop interface, which streams the searches, but it is best-effort: ChatGPT decides which interface it serves, so handle answers without them.
• Web search is forced by default. Leave disableWebSearch unset to measure search-backed answers. With true you get the answer ChatGPT gives on its own, and sources and searchQueries are often empty.

Aggregate across a prompt set

Count each normalized search once per answer that ran it:

from collections import Counter

def fan_out_report(results: list) -> list:
    """Searches with the number of answers that ran them, most frequent first."""
    counts = Counter()
    for result in results:
        counts.update({" ".join(q.lower().split()) for q in result.get("searchQueries", [])})
    return counts.most_common()

results are result objects: from synchronous responses, or response.result of completed tasks.

Turn it into work

• Map searches to pages. For each frequent search, note the page on your site that should answer it, or that none does.
• Join with citations. For each search, collect the domains cited in the answers that ran it. Those are the pages you compete with.
• Split by market. Run the set with each country you sell in; the searches can differ by market.
• Keep the set fixed. Re-run the same prompts on a schedule so changes in the fan-out are visible.

Keep reading