Skip to main content
GlossaryStructured Data

Retrieval-Augmented Generation (RAG)

An architecture where a language model is grounded in real-time retrieved documents instead of training data alone.

Also known asRAG

Every consumer AI assistant in 2026 runs on RAG. When a user asks a question, the system first retrieves passages from a search index, then feeds them to the LLM as context, then asks the LLM to answer citing those passages. GEO is, operationally, about winning the retrieval step.

Understanding RAG clarifies the GEO mental model: you are not ranking for a user — you are ranking for a retriever that feeds a language model. The optimisations differ.

Related terms

All terms
Core GEO

AI Indexing

The process by which AI retrieval systems discover, fetch, parse, and embed a website so they can cite it later.

Shopify Signals

Semantic Chunking

The process a retrieval system uses to break a page into passages for embedding and retrieval.

Content Patterns

Citation Source

A URL or brand that an AI-generated answer attributes as the origin of a statement.