Grounding

The practice of anchoring AI-generated outputs to verified, factual sources—such as your knowledge base, database, or documents. Grounding reduces hallucination by forcing the model to cite real data rather than generating from memory alone. Techniques include RAG, citation requirements, and confidence scoring. Essential for legal, healthcare, and finance agents where accuracy is non-negotiable.