Retrieval-Augmented Generation (RAG)

The short answer

RAG is the process AI search tools use to find relevant web pages for a question, then feed those pages to the language model so it can write an answer with real sources attached.

Full explanation

Instead of answering purely from what it learned in training, a RAG system converts your question into a search, pulls back a handful of matching passages from the web, and asks the model to summarize them with citations. This is why a page has to be crawlable and clearly written to be cited at all — if it's never retrieved, it can never be quoted.

Back to all terms

77 terms and growing