Definition
LLM SEO focuses on the retrieval and citation behavior of large language models. When an LLM answers a question, it either recalls from training data or retrieves fresh sources via a search tool. LLM SEO optimizes for both paths.
The four pillars
LLM SEO rests on four pillars. Weakness in any one caps performance.
- Entities - your brand is a clear, well-described entity across the web
- Structure - content is answer-first, semantically marked up, and easy to extract
- Mentions - you appear in sources LLMs read (Reddit, G2, YouTube, publications)
- Technical accessibility - pages server-render, expose clean HTML, and publish llms.txt
Retrieval vs recall
Recall-based answers come from the model's training data - slow to influence, but persistent once earned. Retrieval-based answers come from live web searches - faster to influence with fresh, well-ranked content. A mature LLM SEO program invests in both.
The role of llms.txt
llms.txt is an emerging convention (analogous to robots.txt) that gives LLMs a curated markdown index of your most important content. It doesn't replace SEO, but it makes selective retrieval easier and signals machine-readability intent.
Measurement
Track prompt coverage, citation share, and the specific pages LLMs cite. Over time, correlate content and PR investments with citation lift.