Grounding is the practice of anchoring an AI answer in retrieved, verifiable sources rather than relying on the model's memorized knowledge alone. A grounded answer can point to the specific pages it used.
In plain terms
Grounding supplies a model with external context, such as search results or retrieved documents, to support a response.
Some providers expose grounding chunks, source links, or citation metadata. Those artifacts show what the provider surfaced, but they do not necessarily map every generated claim to one source.
Grounding can make current information available at answer time, but it does not guarantee factual accuracy.
Why it matters
Grounding can provide current, inspectable context and reduce reliance on model memory alone.
It is evidence support, not a guarantee: retrieved sources can be incomplete, misapplied, or represented inaccurately in the generated answer.