See the searches AI might run before it answers.
One buyer question can fan out into several background searches inside an AI engine. Run any question through Claude, ChatGPT, Gemini, and Perplexity. See the real background search queries Claude, ChatGPT, and Gemini report through their official APIs when they ground an answer, plus the sources and citations Perplexity's API returns, what your content has to win to get cited. When a model answers from its own trained knowledge it runs no live search, and the report marks those rows so you always know which queries were actually captured.
What is GEO Query Fan-Out?
GEO Query Fan-Out is a free, bring-your-own-key tool that takes one question and captures the real background searches Claude, ChatGPT, and Gemini report through their official APIs; and for Perplexity, the sources and citations its API returned when they ground an answer. When a model answers from trained knowledge it runs no search, and the report marks those rows; the captured queries become a content map for what you need to cover to get retrieved and cited by AI engines.
How it works
You type a question (optionally add your domain), paste your own API keys for any of the four engines. Claude, ChatGPT, and Gemini return the actual background search queries they reported when they grounded that question; captured from the response itself, not simulated; Perplexity returns the sources and citations it retrieved (its API does not expose query strings).
Prompts a model answers from its trained knowledge trigger no live search, and those rows are marked accordingly. Add Semrush, SpyFu, or SparkToro keys to layer estimated search-volume, competitor, and audience data over the captured queries; keys travel over HTTPS for the single request and are then discarded, never logged or saved.
Who it’s for
For marketers, SEO and GEO teams, and business owners who want to know which questions to answer to show up in AI answers; the outcome is a concrete list of sub-topics to cover so your content gets retrieved and cited, instead of guessing at keywords. It suits anyone whose buyers now research in AI first and who wants to build that content map themselves before committing to a program.
In practice
A B2B software marketer enters "best CRM for a small business" and pastes their ChatGPT and Gemini keys; the tool returns the background searches each engine ran, things like small-business CRM pricing, CRM vs spreadsheet, easiest CRM to set up, and CRM integrations, revealing that their pages answer the main phrase but ignore the pricing and integrations sub-questions the engines are likely to explore, so they know exactly which pages to write next.
Your question, every engine’s searches and sources.
Add a Claude, OpenAI, or Gemini key to generate your brief. A Perplexity key can be added alongside for retrieval, but Perplexity alone cannot generate the brief (used once for your request, never stored). Add Semrush / SpyFu / SparkToro to layer in estimated search volume, competitive signals, and audience data from third-party providers.
Use your own API keys : used once for your request, never stored on our servers
Keys travel over HTTPS to our server, which calls each AI provider for this one request, then discards them, never logged, saved, or reused. Add any one AI engine to start; add Semrush / SpyFu / SparkToro to compare estimated search data. Want it run for you instead? Talk to us.
Nothing hidden. See every connection.
This is exactly how your report gets built, end to end. Click any node to see what it does and what it feeds, or hit Play for the guided tour. No black box.
Questions, answered.
Yes, the NYFTY tool itself is free to use. Because it is BYOK, any usage charges from the AI provider are tied to your own API key or account.
Yes, this is a bring-your-own-key tool. Your API key is sent over HTTPS to NYFTY's server only to make the request(s) for that single run, is never logged or stored, and is discarded immediately after.
GEO Query Fan-Out captures the real background searches Claude, ChatGPT, and Gemini report through their official APIs; and for Perplexity, the sources and citations its API returned when they ground the same question, so you can compare them side by side. Availability can depend on the API access enabled for the key you provide.
It outputs side-by-side fan-out query lists: the real background searches Claude, ChatGPT, and Gemini reported when they grounded the question (for Perplexity: the sources and citations it retrieved), plus the sources it pulled. Prompts answered from a model's trained knowledge run no search and are marked as such. Use it to spot content gaps, compare how engines interpret a question, and plan pages or sources that better answer those subtopics.
For grounded answers, yes. When Claude, ChatGPT, or Gemini runs a live web search to answer, it reports the background queries it used through its official API, OpenAI's Responses API, Gemini's grounding metadata, Anthropic's web-search tool blocks, and the tool captures those directly. Perplexity is different: its API returns the search results and citations it retrieved, not the underlying query strings. When a model answers from its trained knowledge instead, it runs no search, so there are no queries to show and the report marks that row as no-search rather than inventing one.
AI systems can vary because of model sampling, provider updates, live retrieval differences, and small changes in how a question is interpreted. Treat the output as directional research, not a guaranteed map of future AI answers, citations, rankings, or visibility.