Retrieval-Augmented Generation (RAG)
Updated · By the Everything-PR Editorial Team. Reviewed by a human editor before publication.
Retrieval-augmented generation (RAG) is the technique in which an AI system looks up relevant documents or web pages at the moment a question is asked and gives them to the language model to build its answer. It was first described in a 2020 research paper from Meta AI and is now the backbone of most answer engines.
The flow has four steps: the system interprets the query, retrieves candidate sources from a search index or database, selects the most relevant passages, and then generates an answer that cites them. Because the model answers from retrieved text instead of memory alone, RAG reduces hallucination and keeps answers current beyond the model's training data.
For brands, RAG explains why some pages get cited and others do not. Engines retrieve passages, not whole pages, so retrieval chunking, clear structure and crawl access decide what is eligible. Related concepts: grounding, answer engines and large language models.