What is LLM SEO?
LLM SEO is search optimisation for large language models: making a business easy for ChatGPT, Claude, Gemini and similar tools to know, describe and recommend.
Normal SEO aims for a ranking in a list of results. LLM SEO aims for something else: when someone asks an assistant a question you could answer, it knows you exist, describes you accurately and names you.
It goes by other names too, including AI SEO, GEO and AEO. LLM SEO puts the focus on the models themselves, and on how they come to know anything at all.
How do ChatGPT, Claude and Gemini come to know a business?
Through two routes: what the model learned in training, and what it finds when it searches the web during a conversation. Most buying questions use both.
Training is when a model reads a huge amount of public text and learns patterns from it, including which names go with which topics. If your business was written about clearly and often before the training cut-off, the model may know you without searching. If it wasn't, it won't.
Live search fills the gap. For questions about prices, local providers or recommendations, assistants usually search, read a handful of pages and answer from those. This is how a new business, a new service or a price change can show up soon after the page is crawled.
What's the difference between training data and live search?
Training data is broad but out of date. Live search is current but narrow. You need to show up in both.
- Training data is built from public web text up to a cut-off date. It changes slowly, and shapes what the model already believes about you.
- Live search happens as the assistant answers: it searches, reads the top results and often cites them. It changes quickly, and decides most recommendations for buying questions.
- Training crawlers such as GPTBot and ClaudeBot collect pages for future models. Search crawlers such as OAI-SearchBot and Claude-SearchBot fetch pages for answers. You control them separately in robots.txt.
Why do consistent mentions matter so much?
Because a model trusts what many sources agree on. One clear description, repeated across your site, directories, press and reviews, is far more likely to stick than ten different ones.
If your site says "web design studio", your Google profile says "marketing consultancy" and a directory lists an old address, the model has three competing pictures and no reason to pick any of them. When every source says the same thing in roughly the same words, the picture is clear, and that's the version that gets repeated back.
Mentions without a link count too: press coverage, industry lists, forum threads and podcast show notes. Our clients have been featured in The Times and on BBC Radio 2, and coverage like that is read by people and machines alike.
What is entity clarity, and why does it help?
Entity clarity means a model can tell exactly which business you are, and doesn't confuse you with anyone else. It comes from one name, one description, and links that tie your profiles together.
Models organise what they know around entities: people, places, companies and products. If your business shares a name with another company, or trades under two names, you risk being merged with someone else or left out. These five habits keep you distinct.
- Use one business name, spelled the same way everywhere.
- Write a one-line description of what you do and who for, and reuse it word for word.
- Add Organization or LocalBusiness schema with your name, address, phone, logo and sameAs links to your official profiles.
- Keep an About page that states the facts plainly: who runs the business, where it is and what it does.
- Link from your site to your official profiles, and from those profiles back to your site.
Do ChatGPT, Claude and Gemini search in the same way?
No. Each uses a different search source, so a business can be visible in one and missing from another.
ChatGPT search leans on Bing. Claude's web search uses Brave and other sources. Gemini grounds its answers with Google Search. That's why we set up Bing Webmaster Tools as well as Google Search Console, and check each assistant separately.
Their crawlers have different names too. If robots.txt or bot protection blocks one by accident, that assistant can't read you, even when the others can.
How do you do LLM SEO, step by step?
These are the eight steps we work through, in order. Steps one to four make you easy to know. Steps five to eight make you easy to find today.
- Write your core facts once: name, one-line description, services, areas, "from" prices where you can, turnaround and how to start.
- Put them in plain text on your site, readable without JavaScript, with an About page and one page for each service.
- Add Organization or LocalBusiness schema with sameAs links, plus Service and FAQ schema where they fit.
- Make your name, address, phone and description identical on Google Business Profile, Bing Places and every directory you're listed in.
- Check robots.txt and bot protection let the search crawlers in: OAI-SearchBot, Claude-SearchBot, Googlebot and Bingbot. Decide separately whether to allow training crawlers.
- Get indexed in Bing as well as Google: submit your sitemap to both and turn on IndexNow.
- Earn mentions where models read: trade press, comparison lists, reviews and genuine forum threads.
- Ask the assistants the same buyer questions every month, and log who's named, how you're described and which sources are cited.
How long does LLM SEO take?
Live search can pick up changes soon after a page is crawled again. What a model learned in training changes far more slowly, because it waits for the next model.
So the quick wins are on your site and in the search indexes assistants use. The slow work, mentions and coverage across the web, builds a picture that lasts. BidScript, a UK bid management software company, went from about 10 to about 400 Google AI answers a day in 3 months, on the back of clear facts, comparison pages, structured data and digital PR.