Fan-Out Queries · How-To Guide

How to find ChatGPT's hidden
fan-out queries.

Every time ChatGPT searches the web, it fires off multiple hidden queries behind the scenes. Here's how to uncover them - and why the manual method won't scale.

Ben Tannenbaum · February 15, 2026 · 7 min read

When you ask ChatGPT a question that needs current information, it doesn't just fire off a single search and call it a day. It takes your prompt, breaks it apart into multiple angles, and sends separate web searches for each one. These are called fan-out queries, and they determine which websites end up in ChatGPT's answer - and which get ignored entirely.

What Are Fan-Out Queries?

When ChatGPT searches the web, it generates 8 to 15 separate sub-queries from a single user prompt. Each sub-query targets a different intent, attribute, or comparison angle. The final answer is synthesized from all of them - meaning your brand needs to rank across multiple sub-queries, not just one.

How Fan-Out Works: A Visual Breakdown

Here's what happens behind the scenes when a user asks ChatGPT a single question. The model decomposes it into multiple search angles:

"Best CRM for small businesses?"ChatGPT Fan-OutSUB-QUERY 1"best CRM software smallbusiness 2026 reviews"SUB-QUERY 2"CRM pricing comparisonsmall teams"SUB-QUERY 3"HubSpot vs Salesforcevs Pipedrive"SUB-QUERY 4"CRM features contactmanagement automation"SUB-QUERY 5"best free CRM toolfor startups"10 web results10 web results10 web results10 web results10 web resultsSynthesized Answerfrom 50+ scraped web pages
One user prompt triggers 5+ separate web searches. ChatGPT scrapes and synthesizes results from all of them.

Your website might rank on page one of Google for "best CRM for small business." But if it doesn't show up for the specific sub-queries ChatGPT actually sends, your brand could be completely absent from the AI-generated answer. That's the fan-out gap.

Watch: How Query Fan-Out Works

If you prefer a visual walkthrough, this explainer from Mike King at iPullRank breaks down exactly how query fan-out works across AI search - why one prompt becomes many searches, and what it means for your content strategy:

Query Fan-Out explained: how a single AI search prompt fans out into many web searches. Source: iPullRank.
8–15

hidden sub-queries fired from a single ChatGPT prompt

50+

web pages scraped and synthesized into one answer

1

head keyword you rank for ≠ visibility across every sub-query

The Hard Way: 5 Steps in Browser DevTools

There is a way to see exactly which queries ChatGPT sends when it searches the web. It involves intercepting your browser's network traffic using Developer Tools. Here's the step-by-step process:

1
Send a prompt that triggers web search

Open ChatGPT and type a question that requires current information - something like "What are the best project management tools for remote teams in 2026?" Wait for the full response to load.

Tip: The question must require live data. Generic knowledge questions won't trigger web search.

2
Copy the conversation ID from the URL

In the URL bar you'll see something like chatgpt.com/c/68f1007d-7e08-832b-... Copy the first alphanumeric segment (e.g. 68f1007d). You'll need it as a filter.

Tip: This ID is unique to each conversation and lets you isolate the right network request.

3
Open Developer Tools → Network tab

Right-click anywhere on the page and select "Inspect", or press Ctrl+Shift+I (Windows) / Cmd+Option+I (Mac). Click the Network tab at the top of the DevTools panel.

Tip: If this is your first time in DevTools, it can look overwhelming - but you only need the Network tab.

4
Refresh the page and filter by conversation ID

With the Network tab open, refresh the page. Dozens of requests flood in. Paste the conversation ID into the filter box to isolate the relevant one.

Tip: You're looking for a "fetch" or "xhr" type request in the filtered results.

5
Search the response for search_model_queries

Click the filtered request, open the Response tab, and press Ctrl+F to search for "search_model_queries". If web search was triggered, you'll find the exact queries ChatGPT issued as web searches.

Tip: If the field isn't there, ChatGPT answered from training data alone - no web search happened.

What You'll Find in the JSON Response

If the prompt triggered web search, the JSON payload will contain a search_model_queries field with every query ChatGPT issued - typically between 3 and 15 separate searches.

DevTools - Network - Response
{
  "search_model_queries": [    "best project management tools remote teams 2026",    "Asana vs Monday.com vs ClickUp comparison",    "project management software pricing per user",    "remote team collaboration tools features integrations",    "best free project management tool for startups"  ]}
One user prompt produced five distinct search angles - review content, brand comparisons, pricing, features, and free tiers.

Why the Manual Method Falls Apart

The DevTools trick works for a quick peek, but it completely breaks down the moment you try to use it systematically. Here are the four reasons it won't work for real SEO strategy:

One conversation at a time

Want to analyze 50 product queries? You'll need to repeat the entire 5-step process 50 times.

Queries change every session

Ask the same question tomorrow and ChatGPT may generate completely different sub-queries. A single snapshot tells you nothing about patterns.

Requires DevTools expertise

Not everyone on your marketing team knows how to filter network requests and parse JSON payloads.

No visibility into real user prompts

You can only see fan-out queries for YOUR conversations. Zero visibility into what ChatGPT searches when actual customers ask about your industry.

Browser Extensions Don't Fix the Core Problem

A few Chrome extensions and bookmarklets can automate extracting queries from the JSON response - but they share the same fundamental limitation: they only show fan-out queries for your own conversations. You're still limited to one prompt at a time, one session at a time, with zero visibility into what real users are actually asking about your industry.

Quick Comparison: Manual vs. Aiso

Here's how the manual DevTools approach stacks up against using Aiso's Fan-Out Intelligence:

CapabilityDevTools / ExtensionsAiso
See fan-out queriesYour conversations onlyReal user conversations at scale
Technical skill neededDevTools & JSON parsingNone - dashboard interface
Analyze multiple promptsOne at a time, manuallyHundreds, automatically
Track changes over timeNot possibleBuilt-in time tracking
Competitor visibilityOnly if you test their queriesSee who ranks for each sub-query
Team accessIndividual browser onlyShared dashboards & exports

The Easy Way: See Fan-Out Queries at Scale with Aiso

Instead of manually inspecting your own conversations one by one, Aiso's Fan-Out Intelligence gives you access to fan-out queries generated across real ChatGPT conversations - not just yours, but from actual users asking questions about your industry, your competitors, and your product categories.

Aiso Fan-Out IntelligenceUser: "best CRM for SaaS?"User: "CRM vs spreadsheet?"User: "HubSpot alternatives"User: "cheap CRM free trial"+ hundreds more...AisoAggregates & AnalyzesALL FAN-OUT QUERIESacross your industryCONTENT GAPSwhere competitors rank, you don'tTREND TRACKINGhow queries evolve over timeTEAM DASHBOARDSno DevTools neededReal conversation data from 5M+ ChatGPT conversations
Aiso aggregates fan-out queries from thousands of real ChatGPT conversations - giving you the full picture, not just your own test prompts.

Stop Spinning Up DevTools. Start Seeing the Full Picture.

The DevTools method is useful to prove to yourself that fan-out queries exist. But if you want to use this data strategically - to inform content creation, close visibility gaps, or monitor how AI search evolves for your industry - you need something that works at scale.

That's what Aiso does. No DevTools. No JSON parsing. No spinning up individual conversations one prompt at a time. Just the data you need to make your brand visible where AI search is actually looking.

Frequently Asked Questions About Query Fan-Out

What is query fan-out in ChatGPT?

Query fan-out is the process where ChatGPT takes a single user prompt and breaks it into multiple separate web search queries - typically 8 to 15 - each targeting a different intent, attribute, or comparison angle. ChatGPT then scrapes and synthesizes the results from all of them into one answer.

How many queries does ChatGPT fan out into?

When ChatGPT triggers a web search, it usually generates between 8 and 15 sub-queries from one prompt, though the count varies by how complex and comparison-heavy the question is. Each sub-query returns its own set of web results that feed the final synthesized answer.

How do I find ChatGPT's fan-out queries?

You can reveal them manually with browser DevTools: send a prompt that triggers web search, copy the conversation ID from the URL, open DevTools → Network, refresh and filter by that ID, then search the response for the search_model_queries field. To see fan-out queries across real user conversations at scale instead of just your own, use a tool like Aiso.

What is search_model_queries?

search_model_queries is the field in ChatGPT's network response that contains the exact list of web search queries the model issued for a given prompt. If the field is present, web search was triggered; if it's missing, ChatGPT answered from its training data alone without searching the web.

Does Google AI Mode use query fan-out too?

Yes. Query fan-out is not unique to ChatGPT. Google's AI Mode and AI Overviews, plus Gemini and Perplexity, all decompose a single query into multiple related searches before synthesizing an answer. Optimizing for the sub-queries - not just the head term - is what makes a brand visible across AI search.