RAG (retrieval-augmented generation)
The standard way answer engines work: fetch relevant documents first, then write an answer from them. It is the reason your page content, not the model's training, decides whether you are quoted today.
Updated August 21, 2026
In a RAG system the model does not answer from memory. A retrieval step finds candidate passages, and the model writes using those. That has a practical consequence worth internalising: your pages compete at the retrieval step, before any writing happens. A page the retriever never returns cannot be cited no matter how good it is, which is why accessibility and clear, self-contained passages matter more than polish.
Related terms
- Retrieval. The step where a system finds and fetches your page. If a page is blocked, unreachable, or clearly not about the topic asked, it is never retrieved, and nothing else matters.
- Chunking. The splitting of a page into smaller passages before a machine indexes it. Each passage is judged on its own, so a fact that only makes sense with the rest of the page around it often loses.
- Grounding. Tying a generated answer to real source documents rather than the model's own recollection. Grounded answers carry links, which is where your citation comes from.