Grounding

Level 1

Making sure the AI's answers are based on real facts, not just stuff it made up.

Level 2

Connecting AI responses to real, verified information sources — like search results or documents — so answers are factual instead of invented.

Level 3

The practice of anchoring AI outputs in verifiable external data sources (documents, databases, search results) to reduce hallucination and improve factual accuracy.

Level 4

Conditioning model generation on retrieved evidence or structured knowledge to ensure outputs are faithful to source material — a key technique for reducing confabulation in production systems.

Level 5

Establishing a causal information pathway from verified external sources to generated tokens, ensuring semantic faithfulness via retrieval conditioning, citation generation, and attribution verification — quantifiable through groundedness metrics and source-attribution F1.

Sources

• NIST Trustworthy and Responsible AI Resource Center: Glossary — National Institute of Standards and Technology. Accessed 2026-07-20.