01In plain English
Before the model answers, the tool searches a set of documents, such as your help center or files, and passes the most relevant passages to the model. The answer is then based on that material rather than on the model's memory alone.
02How it works
First, the tool prepares your material. It splits your help articles, contracts or files into short passages and files each one by meaning in a searchable index. When someone asks a question, the tool searches that index and pulls out the few passages that match best. It then hands the question and those passages to the language model with an instruction to answer from them. The model writes a reply in plain language, and the tool can show which passages it used. Imagine a clerk who, before answering a customer, walks to the filing cabinet, pulls three relevant pages and reads from them, rather than answering from memory. When you update a document, the index must be refreshed too.
03Why it matters when you are choosing
Most support bots and "chat with your documents" tools work this way. It makes answers more accurate and lets them cite sources, but they are only as good as the documents behind them.
04What to check
Check which sources it can connect, how often it re-reads them, and whether answers show where they came from.
05Where you will meet it
06Common questions
Does RAG stop hallucinations?
It reduces them but does not end them. The answer is based on retrieved passages, so it is usually closer to the truth and can point to a source. The model can still misread a passage or answer when the search finds nothing useful, so check the cited sources.
What is the difference between RAG and fine-tuning?
RAG looks up your documents at the moment of each question, so answers follow whatever you have stored and update when you change the files. Fine-tuning changes the model itself through extra training. For facts that change, RAG is the usual choice, and fine-tuning suits style and format.
What do I need to set up RAG?
A set of documents that are clean and current, a tool that can index them and a language model to write the answers. Most business products handle the indexing for you. The quality of the answers depends heavily on the quality of the documents, so tidy the sources first.