Guide · LLM SEO
Ranking a page and being named by a language model are two different games. To win the second one, it helps to know how large language models actually come to know about a brand — because there are three distinct paths in, and you optimize for each one differently.
When ChatGPT, Claude, Gemini or Perplexity mentions a brand, that knowledge came from one of three places. Treating them as one thing is the most common LLM-SEO mistake.
Every model is trained on a large crawl of the web up to a cutoff date. If your brand and its facts were widely published and referenced across that crawl, the model may “just know” you with no live lookup at all. You cannot inject yourself into training data on demand, but you influence it the slow way: by being genuinely present and cited across the web over time. This is the long game — and the reason a strong, well-referenced web presence compounds.
Perplexity, Google AI Overviews, and ChatGPT’s search mode do not rely on memory alone. At answer time they run a search, pull in a few current sources, and summarize them — citing as they go. This is the fastest path to appear, because it responds to your current pages, not a year-old training run. Winning it looks a lot like winning search: be crawlable, answer the question clearly, and be a source the retrieval layer already trusts.
When an assistant fetches your actual URL during a conversation, it reads the live page. Here, clean structure and a direct answer near the top decide whether it can extract and quote you. A page that hides its answer behind scripts or clutter is hard to cite even when the model is looking right at it.
You cannot tune what you cannot see. Start with the free AEO / AI-Readiness Checker to confirm the crawlers can reach you, then the AI Visibility Checker (in beta) shows whether the models actually name you — and for which questions. For the strategy behind all of this, see the Answer Engine Optimization guide.
LLM SEO is optimizing so that large language models — the systems behind ChatGPT, Claude, Gemini and Perplexity — know about your brand and cite it in their answers. It overlaps with SEO, but the mechanism is different: instead of ranking a page, you are trying to become part of what the model knows and retrieves.
Three ways, and they matter separately. Training data: what the model absorbed from the web up to its cutoff. Retrieval (RAG): what a search layer pulls in live at answer time, as Perplexity and AI Overviews do. Live browsing: what the assistant fetches from your actual site during a session. You optimize for each differently.
Not directly, and not on demand — but you influence it. Models train on broad web crawls, so being widely published, cited, and referenced across credible sites over time raises the odds your brand and facts are represented. It is a long game measured in mentions across the web, not a switch you flip.
The retrieval and live-browse paths are the fast ones. Make sure the AI crawlers can reach you, answer questions clearly and factually on-page, and earn mentions on sites the retrieval layer already trusts. Those show up in weeks, not model-training cycles.
The next question
Being crawlable is step one. The AI Visibility Checker shows whether ChatGPT, Google AI Overviews, Perplexity and Gemini name and cite your brand — and who they name instead.
Keep going