Fanout Query Extractor
When ChatGPT and other AI engines answer a question, they first split it into a fan of sub-queries. Fanout Query Extractor reconstructs that fan for your topic: two AI models (Perplexity and Gemini) propose the sub-queries independently, the suggestions are merged, and those both models found get the highest confidence. If you provide a reference page, the tool also measures how well your page actually covers each sub-query, using real text embeddings and cosine similarity, not opinion.
The source status in the report is real: you see which models responded and how long they took. The sub-queries themselves are AI reconstruction (two separate, real model calls), the similarity numbers against your reference page are measured, and the insights and H2 suggestions are AI analysis badged accordingly.
Who it's for
Anyone writing content meant to be cited by AI engines who wants to know which sub-queries the article must cover, and where the current text has gaps.
Prerequisites
- Logged-in account with enough credits
- 30 credits per run
- A seed topic of at least 10 characters. Reference URL and title are optional but unlock the gap measurement.
Step by step
- Enter the topic, for example "accountant for sole proprietorships".
- Optionally paste the URL and title of the page you want coverage measured for.
- Pick the market (Norway, Sweden, Denmark, Finland or English) and how many sub-queries you want (5 to 15).
- Run the analysis.
Reading the results
- The confidence badge per sub-query: high means both models proposed it independently; those are the safest.
- The similarity number (with a reference page) is measured semantic similarity: low values are gaps your page does not cover.
- The H2 suggestions are AI-generated headings for the gaps; "Copy all as H2" gives you a ready markdown outline.
- The CSV export takes the whole list to a spreadsheet.
The report is not saved, so export before resetting.
Common problems
- A warning that one source failed: the report then builds on one model instead of two; confidence levels weaken. Consider rerunning.
- Both sources failed: the report comes out empty. Contact support if you were charged for an empty report.
Limitations
- The sub-queries are a reconstruction of how AI engines typically split the topic, not a log from ChatGPT.
- The gap measurement requires a reference URL; without it you get the list only.
- Results are not saved.
Related tools
- AI Consensus, for what the AI engines actually answer on the topic
- Topic Graph Pro, for the broader topic landscape
- Content Brief Generator, to turn the fan into a writing plan
Last updated 2026-08-04