TL;DR: The technique behind AI search where the system retrieves relevant documents before generating an answer, so sources still matter.
In a nutshell
Retrieval-Augmented Generation (RAG) is how Google AI Overviews, ChatGPT search, and Perplexity work. They first search an index for relevant pages, then use those pages as source material when writing the answer. For example, your site getting cited in an AI Overview means it was retrieved as a trustworthy source for that query.
Quick answer: Retrieval-Augmented Generation (RAG) is the technique behind modern AI search. Before generating an answer, the AI system retrieves relevant documents from a search index, then uses those documents as context for its response. This means the answers you see in Google AI Overviews, ChatGPT search, and Perplexity are grounded in real sources, and SEO directly affects which sources get used. Google’s own AI search announcement explains the user-facing experience.
How RAG Works
A RAG system follows three steps:
- Retrieve: The system searches an index for documents relevant to the user’s query.
- Augment: The retrieved documents are added to the AI model’s prompt as context.
- Generate: The AI generates a response, typically with citations back to the retrieved sources.
For website owners, the implication is direct: the better your content ranks in the underlying retrieval step, the more often it gets cited in the generated answer. Topical authority matters more than ever, and crawler rules for AI agents (such as the OpenAI GPTBot docs) are increasingly part of the SEO toolkit.
Where RAG Pulls Its Sources
Modern RAG systems pull from multiple source types:
- Your website (still the foundation, indexed via a strong technical SEO setup)
- Local data (Google Business Profile, citations, the local pack)
- Forums (Reddit especially)
- Reviews (Google reviews, Trustpilot, niche review platforms)
- Q&A sites (Quora, Stack Exchange)
- Video transcripts (YouTube)
- Social posts (X, LinkedIn captions)
- Press coverage and editorial mentions
Brands cited inside AI answers tend to have substantive presence across multiple of those sources, not just on their own websites. This is the core of Answer Engine Optimisation and Generative Engine Optimisation.
Frequently Asked Questions
Is RAG the same as ChatGPT?
No. RAG is a technique. ChatGPT is one product that uses RAG (specifically in its search and browsing features). Google AI Overviews, Perplexity, Microsoft Copilot, and Gemini all use RAG architectures with different implementations.
How does RAG differ from a regular AI chatbot?
Regular chatbots only use what was in their training data. RAG systems retrieve current information from a live search index before generating answers. This makes RAG-powered tools more accurate, more current, and citation-aware.
Can I optimise my site for RAG retrieval?
Yes. RAG retrieval uses search engine indexing under the hood. Strong technical SEO, clear content structure, schema markup, and topical authority all help your content surface during retrieval.
Why is presence beyond my website important for RAG?
Because RAG systems retrieve from multiple source types (forums, reviews, video, social, news). A brand only present on its own website has fewer entry points into the AI’s retrieval set than a brand present across multiple sources.
Take this further
Understanding RAG is foundational to modern SEO. We help clients build presence across the full stack of sources AI tools retrieve from.
Real example: how a Bristol accountancy firm built citation-worthy presence across multiple source types.
Ready to apply this to your own site? Book a free discovery call, or explore our SEO strategy service.