Grounding (AI)
Grounding is the practice of constraining an LLM's response to verified, retrieved sources — the architectural pattern behind every modern AI search engine.
In context
Grounding is the antidote to hallucination: instead of letting the model freely generate, the system retrieves relevant sources first and instructs the model to answer only from those sources, with citations. For SEO/AEO, grounding is why being in the retrieval set matters so much — if your content isn't retrieved, you cannot be cited regardless of how good the model is. Optimisation focuses on being indexable, semantically chunkable, and authoritative for the entity in question.
A Canadian financial advisor's article on TFSA vs RRSP was retrieved into ChatGPT's grounding set for related queries and consistently cited — because the article was crawlable, well-chunked, and topically definitive.
Have a question this didn't answer?
Book a free 60-minute strategy call. We'll explain how this concept applies to your specific situation.