Grounding
Grounding is the practice of constraining an AI model's output to verifiable source evidence rather than its parametric memory. A grounded answer is supported by retrieved passages that can be cited and checked, which is the primary defense against fabricated or confidently wrong responses.
Synonyms: grounded AI, evidence grounding, source grounding, factual grounding
Grounding turns a generative model from a plausible-sounding guesser into an accountable answer engine. By feeding the model retrieved evidence and requiring that its output cite that evidence, grounding keeps answers tied to real, current sources a reviewer can inspect. The discipline extends beyond the prompt: retrieval quality, citation tracking, and a fallback path for missing evidence all contribute. Strong grounding is what makes an automated answer trustworthy enough to resolve a request or trigger a governed action.