This is such a helpful definition of grounding queries vs fan-out queries:

> • *Grounding queries* exist to reduce hallucination and increase reliability, because they pull authoritative signals into the answer-generation process and constrain the model to sources it can confidently reuse, particularly in contexts where factual precision, commercial clarity, or regulatory alignment are required.

> • *Fan-out queries* expand outward from a starting point and allow the AI system to explore adjacent intents before constructing an answer. This is where discovery logic and comparative expansion occur. It is also where AI begins to compress what used to be long research journeys into a single composed response.