Grounding (Search Grounding)

Short answer

Grounding, or search grounding, is the practice of connecting an AI model's answer to retrieved, up-to-date sources, such as live search results or company documents. Instead of relying only on training data, the model bases its response on fetched information. This reduces hallucinations, keeps answers current, and allows the model to cite sources.