How AI answers work
What is retrieval-augmented generation (RAG)?
Retrieval-augmented generation (RAG) is a method that supplies retrieved information to a generative model as context for its answer.
The retrieved material can come from web pages, documents or other collections. Retrieval selects material; generation uses that context to compose a response.
The method can support current or specialised answers, but errors remain possible. Relevant information may be missed or misrepresented.
In Seenford
Seenford measures retrieval-augmented generation outcomes through the answers returned, with exact-sentence receipts for every verdict.
Example
A support assistant retrieves a policy document before answering a question about eligibility.
Related terms
Updated
See what AI says about your brand →