What is LLM SEO?
LLM SEO is the work of making a business easy for large language models to find, read and name. The models behind ChatGPT, Gemini, Claude and Perplexity write one answer to a buyer’s question and name a few businesses in it. Getting into that short list is the goal.
The terms came one after the other and nobody owns them. GEO, AEO, AI SEO and LLM SEO describe the same work with a different emphasis, as GEO vs SEO vs AEO sets out. Generative engine optimization is the most used name.
How is LLM SEO different from classic SEO?
| Aspect | Classic SEO | LLM SEO |
|---|---|---|
| The goal | A position in a list of links | A mention in a written answer |
| The unit that counts | A page | A passage the model can quote |
| Who fetches your pages | Googlebot and Bingbot | Those two, plus the search crawlers of OpenAI, Anthropic and Perplexity |
| Who vouches for you | Links to your site | What other sites say about you: lists, comparisons, trade press, reviews |
| How you measure | Rankings and clicks | How often you are named across repeated runs of fixed questions |
What stays the same: pages that load, text in the HTML and a company that AI systems can identify. If that base is weak, fix it first. The longer version is in AI search optimization.
What do the AI companies say about their crawlers?
Each company runs separate crawlers for separate jobs, and blocking one does not block the others. This is what their own documentation says, read on 27 Oct 2026.
| Company | Search crawler | Training crawler | Fetches on a user’s request |
|---|---|---|---|
| OpenAI | OAI-SearchBot: surfaces sites in ChatGPT search | GPTBot: content that may train its models | ChatGPT-User: robots.txt rules may not apply |
| Anthropic | Claude-SearchBot: improves search results | ClaudeBot: may help train its models | Claude-User: when a user’s question needs the web |
| Perplexity | PerplexityBot: surfaces and links sites; not used for model training | None named | Perplexity-User: generally ignores robots.txt |
| Googlebot: pages must be indexed and snippet-eligible for AI features | n/a | n/a |
Sources: OpenAI, Anthropic, Perplexity and Google. “May” is the vendors’ own wording. Whether to allow the training crawlers is a separate business decision; the search crawlers decide whether you can appear in answers.
What are the six steps?
- Let the search crawlers in. Open yourwebsite.com/robots.txt and check OAI-SearchBot, Claude-SearchBot, PerplexityBot, Googlebot and Bingbot. Check any firewall or bot filter too, which can block them without showing in robots.txt.
- Put the answer in the text. Make the first two sentences of a page answer the question it is about, in plain HTML that needs no JavaScript.
- Make the company identifiable. Ask your web developer to add Organization or LocalBusiness markup: a short block of code that states your company name, address and official profiles (the sameAs list). Google says no special structured data is needed for its AI features; the markup helps other systems tell who you are.
- Use the same facts everywhere. Name, address, phone, prices and description should match on your site and every profile.
- Earn mentions on the sources the engines cite. Ask the engines your buyer questions, note which lists and articles they cite for competitors, and ask each of the three easiest to reach to add or update your listing.
- Measure with fixed questions. Start with 10 questions, 3 times each, in two engines (about 60 runs a month), widen later, and count how often you are named.
The free check covers steps 1 to 3 on your homepage, in part: for step 1 it reads robots.txt but cannot see a firewall or bot filter; for step 2 it checks that the homepage has enough text in its HTML, not that the first sentences answer a question; for step 3 it looks for Organization or LocalBusiness data and linked profiles. It cannot test steps 4 to 6. A tool or a service can run the sixth; best AI SEO tools compares the tools.
Do you need an llms.txt file for LLM SEO?
It is optional. Google’s documentation says you do not need new machine-readable files or AI text files to appear in its AI features, and we found no official statement from OpenAI, Anthropic or Perplexity that their answers use the file. If you add one, it takes about an hour for a small site. What llms.txt is and does covers the details, and the llms.txt generator writes the file for you.
How do you know it is working?
Compare like with like. Keep the question list fixed, run each question 3 times in each engine, and record who is named, in what order, with which sources. One good answer proves nothing, because the same question can give different answers on different runs. The AI visibility audit is that measurement done for you, and AI brand monitoring covers the monthly routine.