Retrieval-augmented generation

Retrieval-augmented generation (RAG) is a method in which an AI system first retrieves documents relevant to a question, then has a large language model generate the answer from them. The term comes from a 2020 paper by Patrick Lewis and colleagues, Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks. It combined a model's built-in, parametric memory with a non-parametric one: a dense vector index of Wikipedia, searched by a neural retriever.

Why it matters for AI visibility

AI search engines follow this pattern. The paper that introduced generative engine optimization describes them as systems that retrieve relevant documents from a database such as the internet, then generate a response grounded on those sources. Google Cloud, likewise, offers grounding in a company's own website data or documents through retrieval-augmented generation. For a brand, this has two consequences. A page can only shape an answer if the engine's search system can reach and index it. And the pages retrieved for a prompt, yours or other sites', decide which brands the answer names.

How to work with it

• Be retrievable. Allow the engines' search crawlers (AI crawler) and keep key facts in the HTML. Google says a page must be indexed and eligible to show with a snippet to appear as a link in AI Overviews or AI Mode.
• Read the citations. The pages an answer cites show what was retrieved and used in that run.
• Answer the follow-up questions. Engines may also retrieve for related queries (query fan-out), so a page that covers pricing, integrations and fit can be retrieved for more of them.

When Perplexity answers "best CRM for a 10-person agency" and cites a review site's category page, that page was retrieved and used. If it lists five competitors and not you, that page is where to start.