Grounding

Tying a model's output to trusted sources, such as your documents, databases or live search, so answers reflect real data rather than the model's memory.

Definition

A grounded answer is one the system can trace back to a source: a retrieved passage, a database record, an API response or a web result. Grounding is achieved through RAG, tool use, semantic layers and search, often with citations shown to the user.

Grounding is the main defence against hallucination in enterprise use, and it is what makes AI answers auditable: a reviewer can check the source rather than trust the model.

Sources

  1. Google Cloud documentation, Grounding overview