Retrieval-Augmented Generation (RAG)

Retrieval-augmented generation, or RAG, combines a generative model with a search step that supplies relevant information at request time. A system converts the query into a searchable representation, retrieves passages from approved sources, and includes them as context for the model’s response. This approach can provide current or organization-specific facts without retraining the model and can support citations. Quality depends on document preparation, indexing, permissions, retrieval ranking, context limits, and the model’s ability to use evidence correctly. RAG reduces some unsupported claims but does not eliminate them: sources may be outdated, relevant material may be missed, and the model may still distort or ignore retrieved information.