RAG retrieves relevant documents first, then lets the LLM answer from those results — helping reduce hallucinations.