RAG (Retrieval-Augmented Generation)
RAG - Retrieval-Augmented Generation - is how modern AI systems answer questions with real sources. Instead of relying only on the model's training data, the system retrieves relevant content from the live web, feeds it to the model as context, and generates an answer grounded in what it retrieved - with citations. When Perplexity or ChatGPT Search answers a question by citing a page, RAG is the pipeline that found your content and decided it was the source.
Why It Matters
RAG is the bridge between classic SEO and AI search. The retrieval step is essentially a search over your content - it favours crawlable pages, clear structure, relevance and authority. If your content is buried, blocked or thin, RAG never retrieves it, and the model answers without you. If it's the best match, you get cited and named.
The flip side matters too: RAG systems are only as good as what they retrieve. That means your page needs to actually answer the query in retrievable form - the answer in the page, not hidden in an image, a PDF, or behind a login. The content that wins RAG retrieval is the content that was already winning search: specific, structured, authoritative and reachable.
In Practice
Keep every answer page crawlable and indexable - the retrieval layer is a crawler, and it can't cite what it can't read. Put the answer in text, early, in the language of the question. Use schema and clean HTML so the system can parse structure. Publish sources and statistics the model can cite. Check whether AI engines actually retrieve you - search yourself in Perplexity and ChatGPT and see whose content gets named.
Treat AI citations like rankings: track who's cited for your industry's questions and audit why. The businesses cited are usually the ones with the most authoritative, most directly answering content - which is the same work as good SEO, with citation-worthiness added.
Common Mistakes
Assuming RAG makes your content irrelevant, or - the opposite - assuming the model can use content it can't retrieve. RAG needs findable, parseable, quotable pages. Hide yours and the answer happens without you.
Sources & Further Reading
Where to verify this yourself - Google's own documentation, industry reporting, and how we apply it at Underdog. Don't take our word for it; check the source.
Related Terms
Glossary
Large Language Model (LLM)
The AI technology behind ChatGPT, Claude, Gemini and the search engines that now read your content. Understanding it is the new SEO literacy.
Glossary
AI Search (Answer Engines)
Perplexity, ChatGPT Search, Gemini - the AI engines people now ask questions to instead of Google. Different rules, same goal: being the cited answer.
Glossary
AI Citation
When an AI answer names your brand or page as its source. The new high-value position - cited beats ranked for the answer itself.
Glossary
Indexing
Adding a crawled page to Google's database so it can appear in search results.
Glossary
Schema Markup
Structured data that helps search engines understand your content and enables rich results.
Know the Words.
Now See Them in Action.
Free teardown. No jargon. Just what's broken and how to fix it.
Get The Teardown