Retrieval, explained
Retrieval-augmented generation, usually shortened to RAG, describes a two-step pattern. First the system retrieves relevant documents, often by running searches against a web index or a private collection. Then the language model generates its answer using those documents as context. Assistants such as Perplexity, ChatGPT with search, Gemini, Copilot and Google's AI Overviews all rely on some form of retrieval.
Retrieval matters to businesses because it is the part of AI search they can influence most directly. A model's built-in knowledge is fixed at its training data cutoff, but retrieved sources are current. If a page is crawlable, fast, relevant and credible, it can be retrieved and cited today, regardless of what the model learned last year.
This is also why SEO remains foundational to AI visibility. Many retrieval systems lean on search indexes, so pages that are not indexed, are blocked to AI crawlers, or load content only through scripts that crawlers cannot run may never be considered. Once retrieved, passages that state facts plainly and near the top of a section are easier for the model to use and attribute accurately.