PLATFORM / FANOUT QUERIES
The Question the User Asks Is Not the Query the Model Runs
A user asks one question; the model runs 8–12 sub-queries behind it and assembles the answer from their results. Fanout Queries captures that invisible tree: which sub-questions you have a source for, where a rival is the only answer, and which are entirely empty — all on one map.
THE CHALLENGE
You Saw One Question. The Model Ran Fifteen.
Optimising for the typed question misses the real work
Models decompose a prompt into sub-queries — price, comparisons, reliability, availability — and answer from those. Your page competes in races you never entered.
Keyword tools cannot see this layer
Most sub-queries are long-tail phrasings nobody types into Google. Zero search volume in a keyword tool; decisive volume inside model reasoning.
Competitors win on branches you never checked
A rival with a mediocre main page can dominate the “is it reliable” branch — and walk away with the recommendation.
Content plans are built on guesswork
Without the sub-query map, editorial calendars chase topics instead of covering the branches that actually feed answers.
HOW IT WORKS
From Invisible Query to Covered Answer
The fanout tree is re-derived weekly for every prompt cluster; as model behaviour shifts, so does the map.
Capture
For every tracked prompt, the searches the model runs behind the scenes and the sources it pulls are recorded.
Tree
Sub-queries are grouped by intent — informational, comparison, trust, commercial — and every branch carries its own volume estimate.
Coverage
Each sub-question is matched against your site: a page that answers it directly, one that touches it, or nothing.
Feed
Empty branches are ranked by impact and sent to Workflow Agent as rewrite or new-page tasks in one click.
THE HIDDEN LAYER
One Prompt, Decomposed
A real unbranded prompt from our coffee-sector report, and the branches a model actually ran behind it. Each branch is a separate contest with its own winners — Fanout Queries shows you every one.
- Real sub-queries per prompt, per engine
- Who appears on each branch today
- Branch-level gaps ranked by opportunity
Budget & price bands
entry vs prosumer pricing
Brand reliability
service network, reviews
Beginner suitability
ease of use, cleaning
Local availability
where to buy, warranty
FROM MAP TO PLAN
Branches Become the Content Calendar
Each uncovered branch is an article brief with intent, the engines that run it and the competitors already cited there. The editorial plan stops being a guess and becomes coverage of a known question space.
- Brief per branch: intent, engines, current winners
- Priority by frequency and commercial weight
- Feeds Workflow Agent for drafting
fanout — branch coverage
WHY IT MATTERS
Eighty Percent of the Answer Comes from Questions You Never Asked
True Query Visibility
Classic keyword research sees the user's question; fanout sees what the model actually searches.
Sub-branch Risk
A sub-branch where a rival is the sole answer can push you out of the main answer entirely.
The Connected Tree
Covering informational branches lifts citation odds on trust branches too — the tree is connected.
Impact-ranked Calendar
The content calendar is built from the impact ranking of empty branches, not guesswork.
SAMPLE TREE
Behind "which Collagen Is Best?"
A shortened example from a real customer cluster; the full tree has 11 branches.
WORKS WITH
From Tree to Action
Fanout data feeds three modules directly.
FAQ
Fanout Queries, Asked Directly
What exactly is query fan-out?
When a user asks a model one question, the model expands it into several focused sub-queries, retrieves for each, then synthesises one answer. That expansion is the fan-out — the layer this module measures.
How do you see the sub-queries?
By observing model behaviour across our daily prompt panel and correlating retrieval patterns per engine — measured, not guessed from keyword data.
How is this different from a keyword tool?
Keyword tools measure what people type into search boxes. Fan-out measures what models ask on their own — mostly phrasings with zero recorded search volume.
Do sub-queries change over time?
Yes, they drift with model updates and season. That volatility is tracked; see our prompt-volatility research on the blog for how much they move.
How does this become a content plan?
Every uncovered branch exports as a brief — intent, engines, current winners — and can be sent straight to Workflow Agent for a draft.
Which engines are covered?
The engines in your plan — ChatGPT, Gemini, Perplexity, Claude and Copilot among them — each with its own branch behaviour.
How does this relate to Prompt Volumes?
Prompt Volumes measures how often user prompts occur; Fanout Queries shows what models do underneath each one. Together they map demand and mechanics.
Which plan includes it?
Fanout Queries ships with Growth and above.
See Where Your Brand Stands Today
The free audit takes about four minutes to set up. The result screen shows your score and the first three fixes.