In grounding, the model is instructed to base its answer on provided documents, search results, or databases and ideally to cite sources. This makes statements traceable and verifiable. Grounding is closely linked to RAG, which feeds exactly these sources into the context. In sensitive areas such as law, medicine, or finance, grounding is crucial for trustworthiness.
Grounding
Grounding means tying AI answers to reliable, verifiable sources instead of letting them be generated freely. It increases factual accuracy and reduces hallucinations.