What Is RAG? Retrieval-Augmented Generation Explained Simply

A language model only knows what it was trained on, and it can't cite a source for what it says. If you ask about your company's internal report, it either doesn't know or makes something up. Retrieval-augmented generation (RAG) fixes this by giving the model the relevant text from your documents at the moment you ask.

1. Your documents are split into small chunks of text.
2. Each chunk is converted into an embedding — a list of numbers that captures its meaning — and stored.
3. When you ask a question, it's embedded too, and the system finds the chunks whose meaning is closest to it.
4. Those chunks are handed to the language model along with your question, and it writes an answer grounded in them.
5. Because the system knows which chunks it used, it can cite them.

The citation is the whole point. It's what lets you verify an answer instead of trusting it blindly, and it's why RAG-based tools are far more useful than a plain chatbot for real work with real documents. See how to do Q&A over your documents with citations for the practical workflow.

Doxy is a RAG application you can use today: upload a document, ask a question, and get an answer with citations back to the exact passages. No setup or machine-learning knowledge required. The same idea of making content legible to AI systems shows up on the publishing side too — see how to validate an llms.txt file if you want AI crawlers to read your own site well.