Free tool · Answer engine optimization
Query Fan-Out Generator
Query fan-out is the set of web searches an AI engine runs before it answers one question. Enter a buyer question below and see the searches a real engine actually issued for it, the terms it added that you never typed, and the sources it read. Nothing is templated: if the engine did not search, the tool says so.
Free, no account required.
What a real fan-out looked like on 4 September 2026
From our own Answer Radar run on the question “Which competitive intelligence tools use Reddit and review site data?”, across four engines. The shape of each engine’s retrieval is different enough to change what you write.
ChatGPT: five shown, eight distinct
"Reddit" "competitive intelligence" tool social listening; Brandwatch competitive intelligence reddit; G2 reviews competitive intelligence software social listening Reddit; Klaviyo competitive intelligence uses reddit; Kompyte Reddit data.
Gemini: two
competitive intelligence tools reddit review site data; social listening sentiment analysis competitive intelligence reddit G2 Capterra.
Claude: one
competitive intelligence tools Reddit review site data. Close to the prompt as typed.
Perplexity: unmeasured
Reports related topics it suggests, not searches it ran, so there is no fan-out to show and Answer Radar says so rather than listing suggestions as searches.
The full measured run, and the exact MCP call that produced it, is on the fan-out product page. The concept, with the research behind it, is in the query fan-out guide.
How to use the query fan-out generator
Step 1
Type a question a buyer would ask an assistant
A full question with a category and a constraint works best: what is the best competitive intelligence tool for a small B2B SaaS team. Bare keywords produce bare answers and often no searches at all.
Step 2
Read the searches, not the answer
The list is what the engine typed into its own search before it wrote a word. Each search is a phrasing a page could answer directly, and a phrasing your buyers never used.
Step 3
Look at the added terms
Years, category words, competitor names and words like pricing or alternatives are what the engine reached for beyond your question. A page whose heading carries them is a page the engine can find.
Step 4
Run it again, then run it on a set
Fan-out changes between runs and between engines. One run is evidence, not a rule. Generate a full prompt set and track it across engines before you plan content around it.
When you outgrow this tool
Answer Radar records the fan-out for every prompt in your set on every run, across four engines, next to who got cited.
This tool is free and complete on its own, with no account and no limits worth mentioning. Linkeddit is what you use when the job stops being a one-off and starts needing to run every week without you.
FAQ
Query Fan-Out Generator questions
What fan-out is, why this tool is not a template, and what one run can and cannot tell you.
What is query fan-out?
Query fan-out is the set of web searches an answer engine issues while answering one question. The buyer types one prompt; the engine turns it into several searches, and those searches decide which pages get read and cited. This tool shows the searches one engine reported for your exact question.
Is this generated from a template?
No. The searches come from the engine's own grounding metadata for your question on this run. If the engine reports no searches, the tool says so rather than inventing a plausible list. Most fan-out generators are templates; this one is the engine.
Which engine does it use?
Google Gemini with Google Search grounding on. Other engines fan out differently: on our own measured run, ChatGPT issued about five searches per answer and named review sites and competitors in them, Gemini about two, Claude one, and Perplexity reports related topics rather than searches. The product view on Answer Radar shows all four.
Is the list complete?
No. A provider may issue searches it does not report, and the same question produces different searches on different runs. Treat the list as evidence of how your question was retrieved against, not as the full set. That caveat is printed with every result because it is true.
Why is there a limit on runs?
A grounded engine call costs more than a plain one, and there is no signup to meter it against. So there is a limit of a few runs a minute and eight a day per visitor, plus an overall daily cap. Repeat questions are cached for a day.
What do I do with the fan-out?
Write or fix the page that answers the recurring phrasing, with the added terms in the heading. Then check whether the engine cites it: the AI visibility checker tests one question, and Answer Radar's content gap view tells you which of your existing pages was supposed to answer and what it is missing.
More free tools
Every tool on the shelf is free and needs no account. See all free tools.
Check whether AI assistants recommend your product for a real buyer question, and see which competitors they name instead.
Generate the buyer prompts AI assistants actually get asked about your category, grouped by funnel stage.
Build a valid llms.txt file for your site, including the citation and safety blocks most generators leave out.
Check whether your robots.txt lets GPTBot, ClaudeBot, PerplexityBot, Google-Extended and the search bots behind them read your site.