GEO Glossary
RAG
RAG (retrieval-augmented generation) means the language model fetches relevant content from a search or database before answering and builds its answer on it.
At a glance
QUESTION │ ├─ 1 Retrieval search on the web or in a database ├─ 2 Chunks matching text passages are selected └─ 3 Generation the model writes the answer from them │ANSWER with sources
Example
The method was described in 2020 in a research paper by Facebook AI Research, now Meta (source: Lewis et al., 2020). Today it is used by ChatGPT with web search, Perplexity and Google’s AI answers.
Why it matters for GEO
This is the only way current content from your website gets into an AI answer. For that your site must be reachable for AI crawlers and have sections that make sense on their own.
In practice
Check whether AI crawlers may fetch your pages and whether important content is in the HTML without JavaScript. Write sections so that each answers a question on its own.
How it fits together
Which documents RAG retrieves is determined by Query Fan-Out. That the answer then actually sticks to these documents is Grounding. You see both as a Citation.