Grounding is the practice of tying an AI response to supplied or retrieved sources so its claims can be checked against real evidence. In a grounded response, the model is not asked to invent an answer from memory alone. It is asked to use information that a person can inspect.
Grounding is useful for two reasons: it can improve factual reliability, and it makes it easier to see where a response came from. A visible AI citation is one practical sign that an answer was grounded in a source.
Is grounding the same as RAG?
They are related but not identical. RAG is a common technical pattern for retrieving information before generation. Grounding is the broader outcome: an answer is anchored to evidence, whether that evidence was retrieved automatically, supplied by a user, or selected by a system.