What is retrieval-augmented generation (RAG)?

Published 14 November 2026. Updated 14 November 2026.

Retrieval-augmented generation (RAG): A method where a language model first retrieves relevant documents, from a search engine or a database, and then writes its answer using them.

The term comes from a 2020 paper by Patrick Lewis and colleagues, which paired a model's built-in knowledge with a searchable index of documents. AI search in ChatGPT, Claude and Grok works the same way, with the live web as the store.

The retrieval step is where fan-out queries are sent and pages are fetched. Fetched vs cited sources covers what happens to those pages next.

Read next

• What is grounding in AI?
• What is an embedding?
• Fetched vs cited sources: why AI answers leave out pages they read

Sources

• Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks (Lewis et al., arXiv, 2020)