Retrieval-Augmented Generation (RAG)

What it is

Retrieval-Augmented Generation is a two-step trick. Before a language model writes an answer, it first retrieves real documents that are relevant to the question, then generates its response grounded in what it found, rather than relying only on what it memorised during training.

The retrieval step usually pulls from a search index. Google's AI features draw from the same core Search index that powers blue links, then layer generation on top. The model also fires off several related queries at once (a technique called query fan-out) to gather a wider set of sources before composing the answer you see.

Why it matters

RAG is the reason your site can be quoted in an AI answer at all. If a page is not in the index, retrieval can never reach it, and generation has nothing to ground itself in. That is why being crawlable, indexable, and genuinely helpful is the whole game for Google's answer engine.

There is no separate ranking trick here. The same clean HTML, clear structure, and trustworthy content that earn ordinary rankings are what get a page retrieved and cited. For the mechanics of how a specific answer chooses which pages to surface, see how AI Overviews pick their sources.