← All terms

Grounding

Anchoring an AI's answer to verifiable source material so it reflects real facts rather than invention.

A grounded answer is one the model can point to a source for. Grounding reduces errors by tying generated text back to retrieved evidence — your page, a document, a database record — instead of letting the model rely on fuzzy memory.

For your business, being a clear, authoritative source makes you good grounding material. The easier it is for a system to verify a fact about you, the more confidently it will state it.

Example

Asked for your opening hours, a grounded assistant cites the exact line on your contact page; an ungrounded one might invent plausible-but-wrong hours.

Why this matters for AI findability

Well-structured, factual content is ideal grounding material. When AI grounds its answers in your site, you get accurate representation and citations — the payoff of agentic findability.