RAG (Retrieval-Augmented Generation)

Retrieval-Augmented Generation is the architecture behind almost every AI search product. Described in a 2020 paper from Facebook AI Research, it pairs a retrieval step — find documents relevant to the question, usually via a search index — with a generation step, in which a language model writes an answer using those documents as context. The point was to give models access to current, specific information they couldn't have memorized, and to let them cite where an answer came from.

For a store, RAG explains why search and AI visibility are one job rather than two. Your page has to pass both steps. In retrieval, it has to rank well enough for the buyer's question to be among the candidates the system pulls — that's search engine optimization, unchanged. In generation, it has to be readable and verifiable enough that the model uses it rather than a competitor's page that was retrieved alongside it — that's the newer work of structured data and plain statements. A page can fail at either step: never retrieved because it doesn't rank, or retrieved and discarded because the facts weren't legible. Checking each step separately (Google for the first, Perplexity for the second) tells you which to fix.