Retrieval-Augmented Generation (RAG)
Retrieval-Augmented Generation (RAG) is a technique where an AI model retrieves relevant external documents or data at the moment of answering a query, rather than relying only on what it learned during training. This lets the model produce more current, source-grounded answers instead of depending entirely on parametric memory.
RAG in Practice
When you ask a chatbot a question about something recent or specific to one company, RAG is often what lets it answer accurately: the system searches a live index for relevant content, feeds that content to the model as context, and the model answers from that retrieved material instead of guessing from training data. This is why Northloom’s returns-policy page — when it’s a clean, indexable, well-chunked web page rather than a scanned PDF linked from a footer — is the version an AI assistant retrieves and cites when someone asks “can I return an opened serum.” The policy itself doesn’t change; its retrievability does.
How to Work With RAG
RAG is a mechanism that runs inside someone else’s AI system — there’s no dashboard where a publisher watches “RAG performance” the way you’d watch a keyword ranking, and it isn’t something you measure directly. What you can actually check is retrievability: is the page indexed, is it broken into self-contained chunks, and does it actually surface when you test your target queries against ChatGPT or Perplexity yourself. Whether your content gets retrieved and cited is really what AI citation rate and citation share (both covered elsewhere in this glossary) measure — RAG is the mechanism behind that number, not a number of its own.
Why RAG Matters
As more search experiences shift to RAG-based retrieval instead of pure parametric recall, content structure becomes a ranking factor in a new way — not just “is this page indexed” but “is this specific chunk of it extractable and citable.”
Terms Related to RAG
Parametric memory ·
Semantic chunking ·
Citation share
Vector embeddings
FAQ
Is RAG the same as fine-tuning?
No — fine-tuning changes the model’s underlying weights during training, while RAG retrieves external information at the moment of the query without altering the model itself.