AI Glossary
Grounding
What is Grounding?
The practice of anchoring AI outputs to verified external sources or real-time data, reducing reliance on the model's internal — and potentially outdated or incorrect — knowledge. RAG is the most common grounding technique.
Example in practice
A financial analyst using an AI tool that pulls live earnings data into its context before responding has implemented grounding — the model's answers are anchored to current facts rather than its training data.