Retrieval-augmented generation (RAG)
Also called RAG, retrieval augmented generation, grounding.
DefinitionRetrieval-augmented generation (RAG) is a method where an AI system retrieves relevant documents or web pages first, then uses them as context to write its answer.
Retrieval-augmented generation (RAG) is a method where an AI system first retrieves relevant documents, such as web pages, then uses them as context to write its answer. Instead of relying only on what the model learned during training, the answer is “grounded” in sources found at the moment the question is asked. The term comes from AI research published in 2020 and is now the basis of most AI search products.
How does RAG work in AI search?
In an AI search engine, the process looks roughly like this:
- The user asks a question.
- The system turns it into one or more search queries (sometimes many, see query fan-out).
- It retrieves the most relevant pages or passages from a search index.
- The language model reads those passages and writes an answer.
- Some of the pages are shown as citations.
Perplexity, Google AI Overviews, Google AI Mode and the search features in ChatGPT, Gemini and Claude all use a version of this.
Why does RAG matter for brands?
RAG is the part of AI visibility you can influence fastest. Changing what a model learned in training takes a long time. Changing which pages a retrieval step finds can happen as soon as pages are crawled and indexed. That puts weight on things marketers already know how to work on:
- pages must be crawlable by the right bots (see AI crawler),
- pages must be findable for the queries the engine runs,
- passages must answer the question clearly enough to be used,
- third-party pages that get retrieved should mention the brand accurately.
Example
A hypothetical solar installer, BrightRoof, launches a new battery storage service. A month later, an assistant answering from training data alone doesn’t know about it. But when the same assistant searches the web, it retrieves BrightRoof’s new service page and a local news article, and names BrightRoof in the answer. The difference is retrieval.
How AI Visibility Desk helps
AI Visibility Desk shows which sources each of six engines cites, including the client’s own cited pages, so you can see what retrieval is picking up. Its site audit checks crawl access for AI bots. Read the LLM SEO guide for more.
Read the full guide
Related terms
Measure it on a real brand
Run a free AI visibility audit across ChatGPT, Gemini, Google AI Overviews, AI Mode, Perplexity and Claude, and get a white-label report.
Start free