Why it matters
Grounded AI answers from approved evidence rather than parametric memory alone. Grounding reduces fiction on factual questions but does not replace governance. Wrong evidence grounds wrong answers confidently.
How it works
Require retrieval hits before answering factual intents, instruct models to abstain when evidence is insufficient, attach citations, and test abstention behavior.
Example
Nintendo chatbot refuses to quote annual plan prices when no verified Pricing FAQ chunk retrieves, linking to the Nintendo Help Center pricing page instead of guessing.
Common mistakes
- 1Grounding labels on answers still allowing uncited claims
- 2Weak abstention when retrieval returns low-confidence noise
- 3Treating grounding as hallucination elimination without tests
Definitions are useful. Governed knowledge is better.
Wiki helps you turn these concepts into a real system for the knowledge behind your AI.
See Wiki in action