Find research angles for your next content brief. Enter a topic to inspect the search queries and sources OpenAI, Gemini or Claude exposes in one API run.
No signup · 20 runs in 24 hoursLOCAL HISTORY
Saved results remain available on this device until you delete them or they are automatically removed after no more than 30 days.
No saved results yet.
0 runs selected
Native fanout shows exposed API searches. Search ideas generates ten modelled research directions without web search.
Choose by the job you need to finish. An API observation and a planning idea lead to different next steps.
You receive the query strings and sources the selected provider exposes in this run. Search actions and metadata help you interpret the observation.
You receive ten modelled research directions without web search. They are a starting point for your own review and remain labelled as ideas.
Both paths start with a short topic. Check demand and evidence next with the Search Console worksheet before turning the questions into a content brief.
For SEO teams, editors and site owners: investigate one concrete question before expanding your content.
When you ask ChatGPT, Gemini or another AI a question and it searches the web, your single question can become a group of narrower searches. “Which SEO tools suit a small business?” might lead to searches about pricing, features, comparisons and user experience. That split is called AI query fanout.
ai-fanout.com runs a bounded web-search request through OpenAI, Gemini or Claude. It then shows only the queries and sources that the selected provider exposes in that run. The tool does not inspect private chats or claim to know every internal search made by a consumer chat interface.
One possible fanout for explanation. A real run may expose different or fewer searches.
Choose country and language to match your market.
You see only queries OpenAI, Gemini or Claude exposes in the run.
A source is connected to a query only when the provider supplies that link.
Keep up to 20 results for no more than 30 days.
The run follows one clear chain: you ask a question, the selected model breaks down the information need, searches the web and shows the search paths it exposed.
Use a keyword or short question of up to 60 characters.
Select OpenAI, Gemini or Claude and add country and language if useful.
The selected model decides which web searches it needs for your topic.
Inspect the exposed queries, cited sources and run details.
The result contains three kinds of information. They belong together but mean different things: the query names the subtopic searched, the source names a website used, and the search action counts the search event.
The tool labels where a result came from. That lets you keep researching when a dataset is missing without inventing demand or hidden provider data.
Query strings and sources exposed by a named provider API in one dated run.
Modelled research directions without web search. Useful for brainstorming, but not observed provider queries.
Use volume, CPC or difficulty only when a keyword source has data for the named market and date. Missing remains unknown, not zero.
A fanout reveals new research paths, but not every query needs its own URL. First identify the user question behind it and check whether an existing page can already answer it well.
Use fanout for SEOWhen the audience, topic and next step stay the same.
When the query has its own user job, evidence and next decision.
When sources are weak or the query is only a hypothesis about demand.
When it is only a wording variant, an outlier or irrelevant to your offer.
The queries reveal possible topics and user questions. The next step is deciding whether they justify a new page, an additional section or no change at all.
Run the same keyword with the same country and language in two selected models. You can then compare which search questions and sources appear in one result or both.
Understand AI fanout comparisonsAn English and a German fanout are not simply translations. The selected country can also bring different products, prices, rules or local providers into focus.
Test country and languageSearch results, available websites and models change over time. Even two nearby runs can therefore expose different searches. Treat each result as a dated snapshot, not a permanent keyword set.
Why fanout results changeFor OpenAI, the web results read by the model also count as input tokens. That is why the input number can be much larger than your keyword. Free use is limited so the tool remains available to everyone.
ai-fanout.com does not permanently store your input or the full result. The visible result stays on your device only when you choose Save. Displayed costs are estimates; the provider's bill is final.
The guides explain fanout queries, sources, model comparisons, changing results and practical SEO decisions. Dated examples show what individual runs actually look like.