Definition
Retrieval-augmented generation
Also known as: RAG
Retrieval-augmented generation (RAG) is an architecture in which a language model is given relevant documents or passages retrieved at query time — from a search index, a vector database or the live web — and asked to answer using them. It grounds the model's output in current, citable sources instead of relying only on what it memorised during training.
Most answer engines are RAG systems: a retriever fetches candidate passages, a reranker orders them, and the generator writes an answer while attributing claims to the passages it used. The term was introduced by Lewis et al. in 2020 (arXiv:2005.11401).
For AEO, the retrieval step is the gate. Content that is not in the index, that is chunked badly, or that does not resemble the question in wording and entities never reaches the model at all.
Frequently asked questions
Why does RAG matter for content creators?
Because it is the mechanism by which your page can be cited. In a RAG pipeline the model only sees the passages that retrieval returned, so a page must first be retrievable — crawlable, chunkable, relevant — before it can be quoted.