# RAG

RAG is when an AI looks things up in your own documents before it answers, so the answer rests on your facts, not just on what the model learned.

Source: https://ibenta.no/en/glossary/rag
Publisher: Ibenta (https://ibenta.no)
Updated: 2026-09-27

---

A [language model](/en/glossary/language-model) knows a lot in general, but it knows nothing about your price list or when the garage closes for the holidays. With RAG, the system first searches a collection of texts you have given it, such as the pages on your website, a price list or a PDF with your terms. The most relevant pieces are sent to the model along with the question, and the model answers from them.

## RAG versus training a model

Training your own model is expensive, takes time and has to be redone with every change. With RAG, you only update the documents. If you change the opening hours on your website, the chat can answer correctly as soon as the new text has been picked up. The answer can also show which page it is based on, so the customer can check for themselves.

It is a common way to build an [AI chat](/en/glossary/ai-chat) that has to answer questions about one particular business. But it is never better than the texts it looks things up in. If the website has the wrong price, the chat repeats the mistake.

## Check your groundwork

Write down the five things customers ask about most often, and find out where the answer is written down today. If it is missing, or only exists in your head, that is where to start. It pays off whether or not you ever get an AI chat, because the same answers help people reading your website.