Retrieval-augmented generation (RAG)

Retrieval-augmented generation (RAG) is the technique most answer engines use to produce grounded, current answers. Instead of relying only on what a model memorized in training, a RAG system first retrieves relevant passages from a search index or the live web, then feeds them to the model to generate a response — often citing the sources it used.

Why it matters for SEO and AEO

RAG is why traditional search signals still matter in the AI era. Answer engines like AI Overviews, ChatGPT search, and Perplexity retrieve largely from existing search indexes, so the same structural signals that help a page rank also decide whether it gets retrieved and cited. Understanding the pipeline clarifies the work: be present and rank in the underlying index (retrieval), and be structured so your passages are the ones selected and quoted (generation). Content that is authoritative, well-chunked, and easy to parse wins at both stages.

Where it applies

• Answer engines — the architecture behind AI Overviews, Perplexity, and ChatGPT search.
• Enterprise AI — internal assistants grounded in a company’s own documents.
• AEO strategy — the mental model for why retrieval (ranking) and extraction (structure) both matter.

What matters from each seat

• In-house — optimize for both stages: rank in the index, and structure passages to be the ones chosen.
• Agency and consulting — explaining the RAG pipeline reframes AEO for clients as an extension of SEO, not a mystery.
• Editorial — authoritative, well-sourced, cleanly-chunked content is what a retrieval step surfaces and a model trusts.