Glossary
Retrieval-augmented generation (RAG)
RAG is a technique where an AI system searches your documents for relevant passages and gives them to the model before it answers, so the answer is grounded in your material.
Why it exists
A model knows what it was trained on, which does not include your price list. RAG closes that gap at answer time: retrieve the relevant passages, put them in front of the model, then generate.
What it fixes and what it does not
It fixes the model not knowing your facts, and it makes answers attributable to a source. It does not fix a model reasoning badly about material it has been given, and it cannot retrieve what you never uploaded.
Grounding is a configuration decision
An agent answering from your documents states your facts. The same agent without them invents plausible ones. That difference is setup, not model quality — which is why "it made something up" is usually a grounding problem.
Related terms
FAQ
What is RAG?
Retrieval-augmented generation: a technique where the system searches your documents for relevant passages and supplies them to the model before it answers, so the reply is grounded in your material rather than in general training data.
Does RAG stop AI making things up?
It substantially reduces it for questions your documents cover, and makes answers attributable to a source. It cannot help with material you never uploaded, and it does not fix a model reasoning badly about text it has been given.