Understand query fan-out
An assistant may decompose a prompt such as “best plumber in London” into related searches about services, location, reviews, alternatives, trust and price. The exact process varies by product and can be opaque, so treat a visible source list or related-query view as evidence to investigate—not a guaranteed transcript of the model’s internal process.
Suggested steps
- Run representative customer prompts in more than one assistant and record the answer, cited sources and any visible related searches or follow-up questions.
- Search the resulting themes in Google, Bing and another relevant search engine. Note whether results come from news, video, community, directories, reviews, professional organisations or competitors.
- Map each theme to a useful page or proof source you can genuinely create. Prioritise questions with clear buying intent and a realistic path to being useful.
- Recheck the sample monthly and after material product, market or search changes. Record date, location, device and prompt so changes are not mistaken for a stable ranking.