Retrieval-augmented generation (RAG) is a method in which an AI system first retrieves relevant documents, often from a search index, and then writes an answer grounded in what it retrieved.
How it works
- A user asks a question.
- The system turns it into one or more search queries.
- It retrieves candidate pages or passages from an index.
- A language model writes an answer from those passages, often with links to the sources.
Google describes its AI Overviews and AI Mode as grounded in its Search index and rooted in its core ranking systems (Google's AI optimization guide). ChatGPT search, Perplexity and Claude run their own search crawlers for the retrieval step (OpenAI, Perplexity, Anthropic).
Why it matters for SEO
Retrieval is a search problem. If a page isn't indexed, can't be crawled, or doesn't clearly answer the question, it won't be retrieved, and a page that isn't retrieved can't be cited. That's why we treat AI visibility as an extension of SEO, not a replacement for it.


RAG is not training
Retrieval happens live, at the moment of the question. Model training is a separate process run by the AI companies on their own schedule and terms. Publishers can allow search bots while blocking training bots, because the companies run them separately.
See also AI crawlers and query fan-out.
