Fanout queries — what AI assistants search behind your answers
See the sub-questions providers looked up internally, and turn 0%-mention queries into content.
Updated 2026-10-12
When an AI assistant answers a question, it rarely searches just that question. It "fans out" into several internal sub-queries — narrower questions it looks up to compose the answer. Maya captures those sub-queries from your tracked prompts. They are the closest thing to seeing the assistant's research notes.
Why fanout matters
Your prompt might be "best romantic hotels in Cappadocia", but the assistant internally searched "cave hotels Göreme price", "Cappadocia hotels with balloon view terrace", "adults-only hotels Ürgüp". Each of those is a retrieval moment where a page gets picked up — or doesn't. If your site never answers the sub-questions, you lose before the answer is even written.
Reading the page
- Stat tiles: Unique queries, Total lookups, Prompts, Avg per run.
- Each row shows the sub-query, how many prompts and providers triggered it, provider logos (ChatGPT, Gemini, Perplexity, Claude, Google AI, DeepSeek, Grok, Bing), and a mention-rate badge.
- A red 0% badge is the signal to act on: providers searched this sub-question and your brand never appeared in the resulting answers. That is a pure content gap with proven retrieval demand.
Filters
Day range, country pills (including "No market"), provider pills, a search box, and a mention filter — All / Mentioned / Not mentioned prompts. The "Not mentioned" view sorted by frequency is the fastest gap list in the product.
Acting on a gap
Every row has "Generate content in Maya Forge" — it seeds an onsite article targeting exactly that sub-question. Fanout query → dedicated answer page is the most direct GEO loop Maya offers: the demand is proven (the assistant already searched it), and the competition is only whatever page it currently picks.
You can also Export CSV to work the list elsewhere.
Notes
- Fanout data accumulates from prompt runs; a new project shows an empty state until enough runs complete.
- Not every provider exposes fanout on every run — counts differ per provider by design.