RAG (retrieval-augmented generation)
RAG is how most AI search works: the platform fetches pages first, then writes its answer from what they say.
RAG stands for retrieval-augmented generation. It's the way most AI search tools answer: first they retrieve some pages, then they generate an answer using what those pages say, on top of what the model already knows.
Without retrieval, a model can only answer from what it learned during training, which can be months old. With retrieval, it can use today's pages, quote prices that changed last week, and link to its sources.
How it works, step by step
- You ask a question.
- The platform searches, often several times, and picks a handful of pages or passages that look relevant.
- It hands those to the model along with your question.
- The model writes an answer, usually leaning on the pages it was given, and may link to some of them.
Think of it as an open-book exam. The model still knows a lot, but for this answer it's reading from the pages in front of it.
Why it matters for your brand
RAG is the reason you can change what AI answers say about you without waiting for the next model to be trained. If a page the platform retrieves names you, you have a good chance of being in the answer. If the pages it retrieves name your competitors and not you, you probably won't be.
For example, if a "best CRM for startups" list is one of the pages a platform retrieves for that question, getting onto that list gives you a real chance of appearing in the answer.
RAG in AnswerOrbit
We keep the pages each answer used and the ones it linked to. Sources shows which pages the platforms retrieve for your questions, and the Gaps tab shows the ones that name your competitors but not you.