The sub-queries AI systems run in the background.
Before an AI system answers a question, it often breaks it down into several sub-queries and searches the web for each one. If you only optimise for the prompt as typed, you miss what the AI is actually searching for.
Sub-queries from real answers.
If a platform discloses its sub-queries, BrandZaps stores them for each answer in their original order. If a platform provides no sub-queries, BrandZaps does not show an empty panel – it shows what was actually measured.
ChatGPT demonstrably discloses its sub-queries.
Measurement in ChatGPT →The measurement ran, but the fan-out stays hidden. This is shown.
Count terms, not phrasings.
Models phrase sub-queries freely every time they run. A ranking of questions would mostly count phrasings and show a one almost everywhere. BrandZaps counts the key terms across all sub-queries.
Coverage checked against your actual website.
The check uses a German full-text search against the title, description and body text of the latest website snapshot.
Without a snapshot, nothing is checked. The view says so explicitly instead of pretending there is a gap.
View website snapshots →Fan-out questions and the domains cited in answers feed back into keyword research and competitor discovery – not an isolated report, but broader research context.
Knowing the original question doesn't mean knowing the fan-out behind it.
The sites that answer the sub-queries are the ones that get cited.




