Free Query Fan-Out Tool: see what AI searches before it answers
Type a question your customers ask an AI. We show the real searches Claude runs to answer it, the pages it reads, and the full fan-out of sub-queries ChatGPT, Gemini and Google AI Mode typically break it into, grouped by type, with the subtopics your content has to cover.
What the Query Fan-Out tool shows you
- Live searches: the exact queries an AI engine sent to the web while researching your prompt, not a guess.
- The pages it read: every result those searches returned, so you see which sites AI retrieves for the topic.
- Your domain check: add your site and see whether it is among the pages AI read.
- The predicted fan-out: 12–18 sub-queries grouped into reformulations, related, comparative, implicit, recency and entity queries.
- The subtopics to cover: concrete content gaps to close so your page is retrieved and cited for the whole fan-out, not just the literal question.
How it works
1. Type a real prompt
Use the question your buyers would ask ChatGPT or Gemini, in their language, not a keyword.
2. AI researches it live
Claude runs the prompt with web search. We capture every query it sends and every page it gets back.
3. See the full fan-out
We group the sub-queries by type and turn them into the subtopics your content needs to cover.
The six types of fan-out sub-queries
AI search engines don't run one search per prompt. They split it into parallel sub-queries of different kinds, retrieve pages for each one and merge them into a single answer. A page that only answers the literal question misses most of them.
Reformulations
The same intent, reworded
Related
Adjacent topics the user likely cares about
Comparative
Head-to-head and alternatives
Implicit
Unstated needs baked into the prompt
Recency
Fresh, time-sensitive angles
Entities
Specific products, brands and features
Why query fan-out decides who gets cited
When someone asks ChatGPT, Gemini or Google AI Mode a question, the engine searches for many sub-queries at once and builds its answer from the pages those searches return. If your page doesn't match any of the sub-queries, it never reaches the answer, however well it ranks for the main keyword.
That's why fan-out is the new keyword research for AI search: the sub-queries tell you which comparisons, constraints, entities and fresh angles your content must cover. Read our guide to query fan-out for the real fan-out we captured across engines, and use this tool to map it for your own prompts.
From one prompt to weekly AI tracking
This tool shows one run of one prompt. SEOcrawl AI's Prompt Tracking runs all your prompts every week in ChatGPT, Gemini, Claude, Perplexity and Google AI Overviews, stores the real fan-out queries of every answer, and shows which pages get cited and whether your brand is mentioned.
Query Fan-Out FAQ
What is query fan-out?
Query fan-out is how AI search engines answer a prompt: they split it into several parallel sub-queries (rewordings, comparisons, implicit needs, fresh angles, specific entities), retrieve pages for each one and synthesise a single answer from them. Google AI Mode, ChatGPT search, Gemini, Perplexity and Claude all do it. Our query fan-out guide explains it in depth.
Are these the real queries AI uses?
The live searches are: they are the exact queries Claude sent to its web search tool for your prompt, captured as it ran. The predicted fan-out is clearly labelled as an estimate: an AI model generates the sub-queries engines like ChatGPT, Gemini and Google AI Mode typically produce for that prompt. Each engine fans out differently, and the same engine can change from one run to the next.
Can I see ChatGPT's or Gemini's real fan-out queries?
Yes, in SEOcrawl AI. Prompt Tracking runs your prompts every week in ChatGPT, Gemini, Claude, Perplexity and Google AI Overviews and stores the fan-out queries of every answer next to the answer, its citations and your brand mentions.
How do I optimise content for query fan-out?
Cover the sub-queries, not just the main question: add the comparisons and alternatives people weigh, answer the unstated constraints (budget, level, use case), name the entities engines search for, and keep time-sensitive facts fresh. Put each answer in its own clearly titled section so it can be retrieved and quoted on its own.
Why does Claude sometimes run no searches?
AI engines decide per prompt whether they need the web. For stable, general knowledge they often answer from memory; for products, prices, recent events or local questions they search. A prompt with no live searches is a prompt where being cited depends on what the model already knows about you.
Is the Query Fan-Out tool free?
Yes, with no sign-up. To keep it fast for everyone there is a daily limit per visitor, and the same prompt is answered from cache for 30 days. To track prompts continuously, try SEOcrawl AI free.
