Skip to content
VectorCite
Free audit

Glossary entry rag

RAG (Retrieval-Augmented Generation)

The architecture every AI answer engine uses: retrieve documents, ground the generated answer in them.

Filed under
Core concepts

Definition

Retrieval-Augmented Generation (RAG) is the architectural pattern where a language model is augmented with a retrieval step. Given a user prompt, the system first fetches relevant documents (via embeddings, BM25, or hybrid retrieval), then synthesizes an answer grounded in those documents with inline citations. Every modern answer engine (Perplexity, ChatGPT search, Claude web tool, AI Overviews) uses RAG.

Example

When you ask Perplexity 'what is RAG', it retrieves ~5 web pages, then writes a paragraph synthesizing them with inline [1][2][3] citations.

How to optimize

Optimize for the retrieval step, not the synthesis step. Your goal: be in the top-5 retrieval cohort. Use BLUF format, high E-E-A-T, schema.org markup, and semantic alignment to buyer queries.

Check a page against a real AI answer

Claude’s live answer to your buyer’s question, its sources, your readiness score and the fixes. Free, no signup, about half a minute.

Run a free audit