Retrieval-Augmented Generation (RAG) is a technique that improves AI answers by first retrieving relevant information from a trusted source, then giving that information to the language model to base its response on. It’s how AI systems answer accurately using specific or up-to-date knowledge.
What it means in plain English
A language model on its own can only draw on what it learned during training, which may be outdated, generic, or simply wrong. RAG fixes this by adding a retrieval step: when you ask a question, the system first searches a knowledge source (often a vector database of documents) for the most relevant passages, then hands those passages to the model and asks it to answer using them. The result is grounded in real, specific information rather than the model’s fuzzy memory.
This dramatically reduces hallucination and lets AI work with private, current, or specialised knowledge it was never trained on — without the expense of retraining the model.
A simple example
An internal company chatbot uses RAG: ask “what’s our refund policy?” and it retrieves the actual policy document, then generates an answer based on it — accurate and current, rather than a plausible guess from the model’s general training.
Why it matters
RAG is one of the most important techniques in practical AI. It’s the standard way to build reliable, knowledge-grounded assistants — from customer support bots to research tools — because it combines a model’s fluency with the accuracy of a trusted source.
Related terms
- Vector Database — stores the knowledge RAG retrieves from.
- Embedding — used to find relevant passages.
- Hallucination — the problem RAG helps prevent.
Frequently asked questions
What is retrieval-augmented generation (RAG)?
RAG combines a language model with a retrieval step: relevant information is fetched from a knowledge source and given to the model so its answer is grounded in that data.
Why use RAG?
It makes responses more accurate and up to date, reduces hallucination, lets AI answer from your own documents, and allows answers to cite their sources.
