G
GEO Toolbox

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

Example, illustrative data:
seed: best ai seo tool
fan-out · 12s
25
queries
19
clusters
4
engines

Real fan-out queries

  • best AI SEO tools compared 2026

    ChatGPT · Gemini · Grok

    Fired2.4K/mo
  • how to measure AI search visibility

    Gemini · Grok

    Fired880/mo
  • GEO tools built for agencies

    Gemini

    Firedparent reach
  • do AI SEO tools actually work

    Perplexity

    Relatedno volume

+ 21 more queries

Divergence 11 shared 8 whitespaceExample data

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.

  1. 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.

  2. 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.

  3. 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.
Intents × engines
fired relatedExample data
Sample cross-engine divergence: which AI engines fan out a query for each intent. A filled dot is a fired search; a ring is a related question (Perplexity).
IntentChatGPTGeminiPerplexityGrok
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
3 shared · 3 whitespace shownRun your own

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 toolsPrompt generatorsQuery Fan-Out
Surfaces the sub-queries AI actually fans outGuessed
Queries are real engine output, not LLM-imaginedSearch dataGuessed
Covers four AI engines in one run1 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 worklistPartial

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.