Concept

RAG (Retrieval-Augmented Generation)

Technique where an LLM retrieves relevant documents or chunks from an external store before generating a response, instead of relying only on training data.

Published February 20, 2026

Definition

RAG (Retrieval-Augmented Generation) is a technique where a language model retrieves relevant documents or text chunks from an external database (or index) before generating a response. The model augments its answer with up-to-date, source-specific content rather than relying solely on its training data.

Production AI assistants (e.g. Perplexity, ChatGPT with browsing) often use RAG: user query → retrieve relevant chunks → inject into context → generate answer with citations. Your website content can feed these pipelines if it is structured for retrieval (clear headings, self-contained sections, JSON-LD, semantic density).

Relevance to GEO

Structuring content for RAG — chunk-friendly sections, FAQ-style Q&A, datePublished/dateModified in JSON-LD, Schema.org types — increases the chance your content is retrieved and cited accurately by RAG-based agents.