Query Fan-Out · 4 engines
AI doesn't answer your keyword.It answers the dozen questions behind it.
Every engine quietly fans one topic into a spray of sub-questions, then answers those. Query Fan-Out captures the real ones — across four engines — maps where the engines diverge, and ranks the content that wins each gap. Real engine output, never guessed.
Free with your own API key · full multi-engine scan from 150 credits in-app
Real fan-out queries
best AI SEO tools compared 2026
ChatGPT · Gemini · Grok
Fired2.4K/mohow to measure AI search visibility
Gemini · Grok
Fired880/moGEO tools built for agencies
Gemini
Firedparent reachdo AI SEO tools actually work
Perplexity
Relatedno volume
+ 21 more queries
The hidden layer
You rank for the keyword and still aren't cited. The reason is invisible in every keyword tool you own: the engine never searched your keyword — it fanned it into ten sharper questions and answered those, citing whoever covered them best.
Query Fan-Out makes that hidden layer visible. It pulls the real questions out of four engines, shows you which ones they all ask and which only one does, and hands you the exact pages that would close the gap.
How it works
From one seed to a ranked worklist.
- 01Seed
One topic or URL
Drop in a seed keyword or a page. Optionally point it at your own URL to score coverage, or a competitor's to see what they already answer.
- 02Fan out
Capture the real queries — across four engines
We read the actual sub-queries each engine fires while answering: Gemini's grounded searches and Grok's many web-search calls carry the depth, ChatGPT contributes its single search call, and Perplexity surfaces its related questions. Real engine output, not an LLM guessing what people might ask.
- 03Rank
Cluster, score, and route to a fix
Queries are clustered into intents, scored by reach and coverage, and turned into a ranked worklist — “write a comparison table answering X,” with the engines that want it and whether you cover it yet.
The divergence map
Where the engines agree — and where only one is looking.
Cluster every fanned query into intents, then lay the four engines side by side. The pattern that emerges is the whole strategy on one screen.
- Shared intents— questions multiple engines fan out. Table stakes: miss one and you're invisible on a query the whole market answers.
- Whitespace— intents only one engine explores. Uncontested: the cheapest citations to win, because nobody's optimizing for a question they can't see.
| Intent | ChatGPT | Gemini | Perplexity | Grok |
|---|---|---|---|---|
| Compare the leading tools | ||||
| How to measure AI visibility | ||||
| Pricing & plans for teams | ||||
| GEO tools for agenciesWhitespace | ||||
| Are these tools worth it?Whitespace | ||||
| Free AI visibility checkerWhitespace |
What's in a scan
Six layers, one run.
The fan-out is the start. Each query is enriched into the full picture you need to decide what to write — and every number is real or honestly labelled, never invented to fill a column.
Validated queries
The questions AI actually fired
Every query is one an engine really searched, tagged FIRED (a real search call) or RELATED (Perplexity's surfaced follow-ups). No “here's what people probably ask” — the honesty is the product.
Divergence map
Where the engines disagree
The signature view: intents shared across multiple engines versus the whitespace only one engine explores. Shared intents are table stakes; single-engine intents are uncontested ground to claim first.
Real volume
Reach, labelled honestly
Each cluster carries true search volume where it exists, parent-term reach for the zero-volume long-tail, and a plain “no volume data” when there's nothing — never an invented number.
Page coverage
What your URL already answers
Point it at a page and every intent is checked against the live content — evidence-verified against the actual text, so a “covered” is a real substring match, not a hopeful guess. WAF-blocked pages are marked unverifiable, never falsely covered.
Citation landscape
Who AI cites on these queries
The sources each engine cites for the fanned queries — you versus the competitors and publishers winning the answer — pulled from the engines' own answer citations, matched at the registrable-domain level.
Ranked actions
A worklist, not a word cloud
Every gap becomes a prioritized action: the format to publish, the intent it answers, the engines that want it, and why it ranks where it does. Start at the top and work down.
How it compares
Validated beats guessed.
| Keyword tools | Prompt generators | Query Fan-Out | |
|---|---|---|---|
| Surfaces the sub-queries AI actually fans out | Guessed | ||
| Queries are real engine output, not LLM-imagined | Search data | Guessed | |
| Covers four AI engines in one run | 1 engine | ||
| Cross-engine divergence map | |||
| Real search volume (honestly labelled) | |||
| Checks your page's coverage of each intent | |||
| Shows who AI cites for the queries | |||
| Outputs a ranked content worklist | Partial |
Run it on your own keys
Don't take our word for it. Watch an engine fan out, live.
The free demo runs a real fan-out in your browser using your own API key. Your key never leaves the page — we don't pay for, proxy, see, or store it. It's the actual engine output, not a recording.
The in-app version adds the engines that can't be called from a browser, real search volume, page coverage, the citation landscape, and saved history — metered at 150 credits a scan, cost shown up front.
FAQ
Frequently asked
- 01What does “query fan-out” actually mean?When you ask an AI engine a question, it rarely answers your exact words. It silently expands your query into a set of more specific sub-questions, searches those, and synthesizes the result. Those hidden sub-questions are the fan-out — and they, not your original keyword, are what decide whether you get cited.
- 02How is this different from a prompt or keyword generator?Keyword tools and “AI prompt” generators ask a model to imagine what people might type. Query Fan-Out reads the queries the engines genuinely fired while answering — Gemini's grounded searches, Grok's web-search calls, ChatGPT's single search call, and Perplexity's related questions. Real engine output beats a plausible guess, especially when you're deciding what to write.
- 03Which engines does it read, and are they all equal?Four: ChatGPT, Gemini, Perplexity, and Grok. We're honest about the differences. Gemini and Grok expose the richest fan-out; Perplexity surfaces related questions (tagged RELATED, not FIRED); ChatGPT's API returns a single search call, so it contributes a baseline rather than depth. Every query shows which engine produced it, so you're never guessing about provenance.
- 04What is the divergence map and why does it matter?It's the grid of intents against engines, showing which questions multiple engines share and which only one explores. Shared intents are the table stakes you must cover to compete at all. Single-engine “whitespace” intents are uncontested — the cheapest citations to win, because most competitors haven't noticed the engine is asking.
- 05Can I run it on a competitor instead of myself?Yes. Seed it with a competitor's brand or point coverage at their URL and you'll see the intents they already answer and the ones they've left open. It doubles as a content-gap map against any domain in your space.
- 06How accurate is the search volume?As accurate as the data allows, and labelled so you always know which you're looking at. Real exact volume when it exists; parent-term reach for the long-tail queries keyword tools score at zero; and an explicit “no volume data” when there's none. We never fabricate a number to fill a cell.
- 07What does it cost to run?A full multi-engine scan is metered at 150 credits ($1.50) in the app, with the cost shown before you run it. Or run the demo free using your own API keys — your keys stay in your browser and we never pay for, see, or store them.
See the questions you're not answering.
Run a fan-out on your topic and start at the top of the worklist.