Retrieval-augmented generation (RAG)

AI search basicsAlso called RAG, retrieval

Retrieval-augmented generation is when an AI assistant first looks up passages from outside sources, such as web pages, and then writes its answer from what it found instead of relying only on memory.

Why it matters

This is why your website can change what an assistant says about you. When an assistant searches before it answers, it isn't judging your whole page the way a search ranking would. It's pulling out the passage that answers the question and quoting from that.

So the plain-text facts on your pages matter more than the page's overall polish. Our post calls the step that does the picking the retriever, and if the answer a customer wants sits in one clear paragraph, that paragraph is what it can pick up.

Example

A homeowner asks an assistant what an emergency drain call costs in her town. The assistant searches, pulls a short paragraph from a plumber's price page that states the call-out fee and the service area, and writes its answer from that paragraph. A competitor with a nicer site but no prices in plain text gives it nothing to quote. The fix is cheap. Put the price, the area you serve, and your phone number on the page as text, under a heading that matches the question.

Ready to see how AI describes your business?

Set up in ten minutes with a 14-day free trial on every self-serve plan, or talk to a specialist first.

Start Free Trial