Retrieval-augmented generation
Also called RAG, retrieval augmented generation.
RetrievalRetrieval is finding and returning information relevant to a request from stored or connected material, so an AI app can use it as context.-augmented generation, or RAG, finds relevant source material and gives it to an AI modelAn AI model is the trained system an application uses to make predictions or generate responses. before the model writes its answer.
Example
A volunteer asks how to cancel a booking. The system retrieves the current cancellation section, then draftsA draft is work that is not finalized yet. an answer with a link to that section.
- A question“How do I cancel a booking?”
- Search the sourcesHandbook, files, earlier chats.
- Relevant passagesThe cancellation section, not the whole handbook.
- Write the answerThe model uses those passages and links to them.
Why it matters
RAG lets answers use information beyond the model's training, such as your own documents, while the source stays separate from the model. The name comes from a 2020 research paper that paired a retriever with a text generator.
The answer is only as good as what was found. Wrong, stale, or missed material can still produce a bad answer.
Common confusion
RAG does not retrain the model and does not guarantee the answer is true. A citation is only useful if it actually supports the sentence beside it.
Sources
Updated October 1, 2026.