Grounding
Grounding links an AI answer to verifiable evidence. A source link alone does not show that every claim is supported.
Grounding links an AI answer to a traceable basis, such as supplied documents or other verifiable data. The answer should rely on that evidence and respect its limits.
A delivery list example
A hypothetical checklist states that a power adapter is included. The assistant can answer that point from it. The list does not, however, establish Friday delivery; that addition would be unsupported.
RAG is one way of supplying suitable evidence. Grounding is not identical to search itself. Even faithfully reproducing an outdated source can be wrong for today's task. Faithfulness, validity and factual quality remain separate checks.
Sources
Related terms
- Retrieval-augmented generationRAG combines information retrieval with generated answers. Access to sources does not remove the need for quality and permission design.
- AI hallucinationAn AI hallucination is plausible-looking content that is factually false or invented. Fluent language is not evidence of correctness.
- LLM evaluationLLM evaluation tests model outputs against defined tasks and criteria. One good example does not establish overall quality.
- Model cardA model card documents an AI model’s purpose, foundations, evaluation conditions and known limitations for a particular version.
- AI readinessAI readiness concerns the foundations of an AI project: objectives, usable data, skills, responsibilities and an operating workflow.
- Context engineeringContext engineering prepares information for individual model calls: relevant instructions, messages, sources and current states.
Articles on this topic
AI & Automation with ASES
AI assistants and automated workflows