Glossary
What Is Retrieval-Augmented Generation (RAG)? Definition
Retrieval-augmented generation, or RAG, is a method where an AI first retrieves relevant documents and then uses them to write its answer, so the reply is grounded in specific sources.
What it means
Instead of relying only on what it learned in training, the system searches a set of documents, such as your company files, and gives the best passages to the model. This usually reduces made-up answers and lets the tool cite sources, but it does not remove errors completely.
Example
A support assistant answers a question by first finding the right paragraph in the product manual, then summarising it.
Why it matters
RAG is behind many tools that answer questions about your own documents, and it is a practical way to improve reliability.
Related terms
- AI Hallucination: An AI hallucination is a confident answer from an AI that is not true, such as a made-up fact, number, quote.
- Large Language Model (LLM): A large language model, or LLM, is an AI system trained on huge amounts of text to understand and generate.
- Context Window: A context window is the amount of text an AI model can consider at once, including your prompt, any documents.
Learn this live, for free
Join the free 90-minute live masterclass, every Friday, 7 PM IST (India) or 7 PM ET (everywhere else). No tech background needed.
Register for the masterclass nowEducation only. This glossary is general information, not financial, legal or career advice.
Frequently asked questions
Does RAG stop hallucinations?
It reduces them by grounding answers in sources, but you should still verify.
Do I need RAG to use AI at work?
No. Many tools do it for you when you upload documents.
What is an example of RAG?
A tool that answers questions only from the PDFs you upload and shows where each answer came from.