Retrieval-Augmented Generation (RAG)

Retrieval-Augmented Generation (RAG) is an architecture in which a language model first retrieves relevant documents — from a search index, vector database, or specific site — and then generates its answer using them as context. RAG reduces hallucination and lets models cite material newer than their training cutoff. AI search engines like Perplexity and ChatGPT search are, in essence, RAG systems running over web-scale indexes.