Generation & Quality

What Is Grounding in AI Answers?

Tying an AI answer to specific source content, so each claim comes from retrieved passages rather than the model's memory and can be checked against them.

A grounded answer is built from evidence the system can point to. Instead of answering from training data, a RAG system retrieves passages from a trusted source and asks the model to answer from those passages only.

Retrieval alone doesn't guarantee it. The model can still add a default value or a step that appears nowhere in the passages, and those additions read exactly like the grounded parts. Stronger systems check each claim against the passages and treat an unsupported claim as a failure. Citations let readers make the same check.

Grounding also limits what an assistant should say. If the retrieved content doesn't cover the question, there is nothing to ground an answer in, and the honest output is abstention. See citations and grounding for how Kelu does this end to end.

In Kelu

How Kelu uses this

Kelu answers only from passages retrieved from your knowledge base and returns a structured citation for each source the answer uses. Where a wrong answer costs more, on the support form and in helpdesk automation, a second model call also checks the answer against its passages. An answer with any unsupported claim gets zero confidence, so it doesn't deflect a form or go out on a ticket.

From definition to deployment

Connect a knowledge source and see cited AI answers on your own content in minutes — free, no credit card required.

Free to start · No credit card required