Grounding

Definition

Tying an AI model's answer to verifiable external sources so its claims can be attributed and trusted.

Grounding is the practice of anchoring a generated response in retrieved documents rather than the model's parametric memory, reducing fabrication and enabling citations. A grounded engine fetches sources, uses them to compose the answer and links back to them, which is why crawlable, unambiguous content is more likely to be quoted accurately. For brands, being the kind of source engines choose to ground on - clear, corroborated and authoritative - is the core of earning reliable AI citations.