Short Definition
Retrieval-augmented generation combines search or retrieval with a generative model so answers can use external context.
Intuition
The system first finds relevant material, then asks the model to answer using that material.
Technical Definition
RAG architectures retrieve documents or passages at query time and condition generation on the retrieved context.
Example
A support bot retrieves product documentation before answering a customer's question.
Common Misunderstandings
RAG does not automatically prevent hallucination.
Retrieval quality often matters as much as model choice.