Retrieval-augmented generation (RAG)
A way of building an AI answer in which the system first retrieves relevant documents, such as web pages, and then asks a language model to write the answer from them. Most AI search features work this way.
Last updated by Vouch
Example: when ChatGPT search or Perplexity answers a question about current prices, it fetches pages first and writes from what it found, which is why it can show citations. The technique was described in a 2020 research paper by researchers at Facebook AI Research, University College London and New York University.
For brands, retrieval is the step where your pages are either found or missed. See grounding and how Perplexity chooses sources.
More terms
See every term in the glossary, try the free GEO tools, compare AI visibility tools, or read the Vouch methodology and pricing.