RAG (Retrieval-Augmented Generation) is a technical architecture used by many AI systems (including AI Overviews and conversational chatbots) in which the language model doesn’t generate answers solely from knowledge learned during training, but first retrieves relevant, recent information from an external source (such as the web search index) and uses it as context to construct the final answer.
This architecture explains why visibility in the traditional search index remains relevant for AI Search: RAG systems often cite the retrieved pages directly, which is why optimizing for content crawlability and extractability matters just as much as for classic SEO.
Why it matters
Understanding RAG helps explain why well-structured, up-to-date, easily extractable content is more likely to be cited by AI Overviews, ChatGPT, or Perplexity, compared to ambiguous or poorly organized content.