Query fan out is what Google AI Mode does instead of running your search. It reads the question, works out the several things it would need to know, searches for each of them separately, and writes one answer from everything that comes back.
Classic search matches a keyword to a page and ranks the results. One search, one list. AI Mode runs several searches the person never typed, then cites a handful of pages from across all of them. The question is the input. The searches are Google's to choose.
Someone asks AI Mode "best CRM for a 5-person marketing agency". The answer cited 20 sources. Reaching those sources takes 5 different searches, and only one of them looks like the original question.
Nobody typed "CRM for virtual assistants". It reaches a source the answer used anyway.
A question like that one carries several buried questions: what agencies need, what small teams can afford, how the main options compare. Fanning out lets AI Mode answer all of them in one pass instead of returning a list and leaving the work to the reader.
Practitioners file this under generative engine optimization, alongside AI Overviews and assistant citations.