AI Glossary
Grounding
Connecting AI outputs to verifiable facts
Definition
Grounding is the process of connecting AI model outputs to real-world facts, documents, or data sources to improve accuracy and reduce hallucination. A grounded model cites its sources, and its claims can be verified. RAG is the most common grounding technique — the model must base its answer on retrieved documents rather than memory alone.