Definition
What is retrieval-augmented generation (rag)?
Retrieval-augmented generation (RAG) is a pattern in which relevant passages are retrieved from a document or data store and supplied to a language model as context before it generates an answer. It grounds output in specific, current, owned content rather than training data alone.
Updated 3 September 2026
Why it matters
- It is how a model answers questions about your policies, contracts or records.
- It allows citation of sources, which makes review and audit practical.
- Most quality failures attributed to models are actually retrieval failures.
Business example
A support tool retrieves the three most relevant policy passages for a customer question and drafts a reply citing them, so the agent can verify the answer in seconds.
Common misconceptions
- 'RAG eliminates hallucination.' It reduces it substantially; grounding plus citation plus review is what makes it safe.
- 'Index everything.' Indexing outdated or duplicated content is the most common cause of wrong answers.
- 'It is a one-off build.' Content freshness and retrieval quality need ongoing ownership.
Related concepts
- Large language model
- Generative AI
- AI governance
Where would this apply in your business?
Merjora maps your workflows and quantifies the opportunities worth acting on.
Discover your AI opportunitiesRelated reading
Editorial standard. Merjora publishes analysis, frameworks and publicly documented examples. We do not publish invented statistics, unattributed benchmarks or unverified customer stories.