RAG (Retrieval-Augmented Generation)
A technique where the system first looks up relevant real documents, then hands them to the model so its answer is grounded in actual sources.
It is the main defense against hallucination and how AI answers questions about your own private data.
Frequently asked questions
What does RAG do in simple terms?
RAG, short for retrieval-augmented generation, first looks up relevant real documents and then hands them to the model so its answer is based on actual sources instead of memory alone. It is the difference between the AI guessing and the AI reading before it answers.
How can AI answer questions about my own company documents?
RAG is the usual way: your documents are stored and searched, and the most relevant ones are fed to the model at answer time so it can respond about your private data. The model itself was never trained on your files; it is handed them on the fly.
Does RAG fix the problem of AI making things up?
It is the main defense against hallucination because the answer is grounded in retrieved sources you can check, but it is not a perfect guarantee. Good RAG greatly reduces made-up answers and lets you trace where an answer came from.
New to all this? Start with what an AI agent really is, browse the full glossary, or explore the learning hub.
← Back to the glossary