What is query fan-out? The searches ChatGPT, Perplexity and Copilot run before they answer

Query fan-out is the set of web searches an AI assistant runs to answer one question. Real fan-out queries from ChatGPT and Copilot, what they add to a question, and how to collect them as GEO data.

What query fan-out is

An AI assistant that answers from the web searches first. One question becomes one or more queries, each query returns pages, and the assistant writes its answer from those pages and cites some of them. The set of queries is the fan-out.

Google uses the name for AI Mode, which breaks a question into subtopics and searches for each. ChatGPT, Perplexity and Microsoft Copilot search the same way, and they show the queries they ran. For GEO the queries matter because they decide which pages the model reads before it writes, and so which pages it can cite.

Real fan-out queries

Four answers collected with querying.ai between September 10 and October 3, 2026, with the searches each engine ran:

Three things happen in these rewrites.

• The engine adds what it expects the answer to contain. ChatGPT put Sony, Bose and Apple into its search, and both WIRED pages it cited name Bose, Sony and Apple in their titles. Brands the model already links to a question shape the pages it gets to read.
• Attributes and freshness join the query. Both engines carried over what the question asked about, noise cancellation and battery life, and Copilot added 2026.
• The place goes into the words. From SoHo, ChatGPT kept “near me” and added what a laptop worker needs, wifi and outlets. From Wangfujing it named the city and the district and asked for opening hours.

What fan-out means for GEO

• Write for the searches. A page gets cited after a fan-out query finds it, so the words engines add, such as brand names, the year, features and the neighborhood, belong in the page's title, headings and text.
• Be one of the brands in the query. When the engine adds three brands to a buying question, pages about those brands are the ones it finds. Track how often your brand appears in the searches as well as in the answers.
• Check each engine. ChatGPT ran one search for the earbud question and Copilot ran two, with different words, so each engine reached its own set of pages. Our citation analysis found that 88% of the websites cited for a question appeared on one engine only.
• See what the searches return. Fan-out queries are ordinary search queries, so their search results show the pages the engine chose from. Run them through search results in the same country.

Collect the searches as data

querying.ai returns the searches as searchQueries next to the answer and its sources. ChatGPT and Copilot return them on every answer; on Perplexity, set include.searchQueries to true. Gemini, Google AI Overviews, Google AI Mode and Naver return the sources they cite, which show where their searches landed.

curl -X POST https://api.querying.ai/v1/async/task \
  -H "Authorization: Bearer $QUERYING_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "taskType": "CHATGPT",
    "payload": {
      "prompt": "Search the web and recommend 3 wireless earbuds for commuting. Compare noise cancellation and battery life in a short table, with source links. Keep the answer under 180 words.",
      "country": "US"
    }
  }'

The finished task carries the searches with the answer, trimmed here:

{
  "response": {
    "text": "For commuting, I’d prioritize strong ANC (trains, buses, traffic) and 6+ hours of playback. …",
    "searchQueries": [
      "best wireless earbuds commuting noise cancellation battery life Sony Bose Apple"
    ],
    "sources": [
      {
        "position": 1,
        "url": "https://www.wired.com/gallery/best-noise-canceling-earbuds/",
        "label": "Best Noise-Canceling Earbuds: Bose, Sony, Apple, and More | WIRED"
      },
      {
        "position": 2,
        "url": "https://www.tomsguide.com/audio/earbuds/best-noise-canceling-earbuds",
        "label": "Best noise-canceling earbuds in 2026: Tested and reviewed | Tom's Guide"
      },
      {
        "position": 3,
        "url": "https://www.wired.com/story/best-wireless-earbuds/",
        "label": "Best Wireless Earbuds We’d Buy Right Now (2026): Apple, Sony, Bose, and More | WIRED"
      }
    ]
  }
}

A fan-out study for your category takes five steps:

1. List the buyer questions you want to win.
2. Run each question on ChatGPT, Perplexity and Copilot in every country you sell in, a few times each, with Answers.
3. Collect searchQueries and group the words the engines add: brands, features, years and places.
4. Compare those words with your pages' titles and headings, and with the pages the answers cite.
5. Track the questions over time with a monitor.