RAG.

Retrieval-augmented generation combines an information-retrieval step with a generative model. The system finds relevant material and supplies it as context for the model’s response.

A typical pipeline prepares documents, searches them for passages relevant to a query, and includes selected passages in the prompt. Retrieval may use keyword search, embeddings, or a combination. More advanced systems can rerank results or perform additional searches when information is missing.

RAG can make private or current information available without changing the model’s weights. Its quality depends on the documents, retrieval method, and the model’s use of the evidence. A missing document, poor search result, or incorrect interpretation can still produce a bad answer.

Retrieval does not automatically make an answer true. The source can be wrong or outdated, and the model can go beyond what it supports. A citation is useful only when it actually backs the associated claim.