Simple definition
Grounding connects AI responses to relevant facts or retrieved sources to improve factual reliability.
Purpose and use
Grounding connects a model's answer to reference information supplied or retrieved for the request.
How does it work?
The system adds relevant passages, tool results or data to context to better support generated claims.
Example and verification
Evaluate whether cited sources actually support important claims and whether those sources are current enough.
Points to watch
Retrieved material can be incorrect, outdated or unauthorized. Grounding reduces selected uncertainties without guaranteeing freedom from invention or misinterpretation.
In short
Quality depends on selecting suitable references and using them faithfully in the answer.