What Is LLM SEO?
LLM SEO is useful as a label only if you separate three things that are often mixed together: what a model remembers, what it looks up, and what it opens when asked.
Short answer
LLM SEO is a label for influencing what large language models say about your brand. It spans 3 layers: training data, search indexes read at answer time, and pages fetched on a user's request. Each layer has different bots and controls, and only the last two decide whether you are cited. None offers a guaranteed mention.
One label, three layers
Language models learn about your business from three different places. Mixing them up is the source of most confused advice on this topic, because a tactic that works on one layer does nothing for another.
| Layer | What it is | Bots involved | How quickly it responds |
|---|---|---|---|
| Training data | What the model absorbed before release | GPTBot, ClaudeBot, Google-Extended | Slow, tied to model releases |
| Search index at answer time | Pages an assistant retrieves while writing | OAI-SearchBot, Claude-SearchBot, PerplexityBot, Googlebot, Bingbot | Days, depending on crawling |
| Live fetch | A page opened because a user asked | ChatGPT-User, Claude-User, Perplexity-User | Immediate, when asked |
"LLM SEO" is a name for working on those layers deliberately. It overlaps with GEO and AEO, and in our usage it is the term to reach for when the question is about the model, not the page.
What each company documents
The three AI companies publish crawler pages. We read them in November 2026.
- OpenAI. OAI-SearchBot surfaces websites in ChatGPT's search features. GPTBot crawls content that may be used to train its foundation models. ChatGPT-User handles user-initiated actions, and robots.txt rules may not apply to it. Each setting is independent, and changes can take about 24 hours to affect search.
- Anthropic. ClaudeBot collects content that could contribute to model training. Claude-SearchBot indexes content to improve search results, and blocking it may reduce visibility there. Claude-User retrieves pages when a user's question requires it. Anthropic says its bots honour robots.txt and supports the Crawl-delay extension.
- Perplexity. PerplexityBot surfaces and links sites in search results and is not used to crawl for foundation model training. Perplexity-User visits pages to answer a user's question and generally ignores robots.txt because a person requested the fetch.
- Google. Google-Extended is a robots.txt token that lets publishers manage whether content may be used for training future Gemini models and for grounding in Gemini Apps and the Vertex AI API. Google says it does not affect inclusion in Google Search or act as a ranking signal. Google also says it intends to support a shared AI training preference signal called AIPREF in early 2027.
What you can influence in each layer
Training data. Very little, and slowly. Your real choice is permission: allow or refuse use of your content for training. Neither choice is documented to change whether search features cite you. If a model already "knows" your brand, that comes from what was public and widely repeated when it was trained.
Search index. Most of the practical work lives here. Keep pages crawlable and indexed, allow the relevant search bots, and write passages that answer questions directly. Google's guide and the 2023 GEO research both favour specific, sourced content.
Live fetch. Make key pages load without a login, with the content in plain HTML. Because robots.txt may not apply to these requests, a firewall rule is the only block, and an overzealous one can hide you from users who paste your link into an assistant.
A robots.txt grid
| Your position | What to allow | What to disallow |
|---|---|---|
| Want AI visibility, comfortable with training | All named search bots and training bots | Nothing |
| Want AI visibility, object to training | OAI-SearchBot, Claude-SearchBot, PerplexityBot | GPTBot, ClaudeBot, and Google-Extended if you accept the Gemini trade-off |
| Want no AI use at all | Nothing | All of the above, plus a firewall check |
| Unsure | Search bots now | Revisit training bots when you have decided |
For the second row the file would look like this:
User-agent: OAI-SearchBot
Allow: /
User-agent: Claude-SearchBot
Allow: /
User-agent: PerplexityBot
Allow: /
User-agent: GPTBot
Disallow: /
User-agent: ClaudeBot
Disallow: /
Test the result in your server logs after a few days, because a CDN or security plugin can override robots.txt with its own rules. Also note the Gemini trade-off: Google says Google-Extended covers grounding for Gemini Apps as well as training, so blocking it is a different decision from blocking a training-only bot.
A diagnostic: which layer is your problem?
Match the symptom to the layer before you spend anything.
| Symptom | Layer to look at | First check |
|---|---|---|
| An assistant describes your company wrongly with no web search shown | Training data and memory | What is public and widely repeated about you, and whether old facts are still circulating |
| An assistant with search on never names you, though you rank in Google | Search index | Whether its search bot can fetch your pages, and whether passages answer the question |
| An assistant cannot read a link you paste | Live fetch | Login walls, scripts, geographic blocks and firewall rules |
| An assistant names you but gets the price wrong | Search index or live fetch | The wording and dating of your pricing page, and old copies elsewhere |
| You appear in Google AI Overviews but not in ChatGPT answers | Different systems | Google's controls and OpenAI's are separate, so check each bot's access |
The first row has the slowest cure. You cannot edit a trained model. You can publish clear, consistent, dated facts and wait for the next release, and meanwhile make sure the search layers carry the correct version.
Timescales, so you can plan
Changes reach the layers at different speeds. OpenAI says robots.txt changes can take about 24 hours to affect its search results. Search index changes depend on when each bot next visits. Training changes depend on release cycles that the companies do not publish. Plan monthly measurement for the first two layers and treat the third as background.
What LLM SEO cannot do
- It cannot make a model say something accurate about you. Models can be wrong, and the fix is to publish clear, consistent facts, then monitor.
- It cannot guarantee a citation. Google's guide says indexing and serving are not guaranteed, and the other companies publish no guarantees either.
- It cannot be verified by tools with inside knowledge. Google states that no third-party tool has access to its internal ranking or AI systems.
Next steps
Work out which layer your goal sits in. If it is being named in answers, read how to rank in ChatGPT for the practical plan. If you are comparing labels, GEO vs SEO vs AEO shows what differs. And if someone has offered you an llms.txt file as a fix, see what llms.txt is and whether you need one. Our AI search optimization service includes a crawler access check. Open your robots.txt today and check that the search bots are not blocked.
Sources
Frequently asked questions
How to do LLM SEO?
Separate the layers. For retrieval, keep pages indexable and allow search bots. For live fetches, make sure pages open without a login. For training, decide whether you want your content used at all. Then write specific, sourced passages and sample the results monthly.
What is LLM SEO called?
It goes by GEO, AEO, LLMO and AI SEO as well. The names overlap and none has a formal definition. Google treats optimizing for its generative features as part of SEO, so the label matters less than the layer you are working on.
Does blocking GPTBot remove me from ChatGPT?
Not from ChatGPT search. OpenAI documents GPTBot as the training crawler and OAI-SearchBot as the one that surfaces sites in search, and says the two settings are independent. Block the search bot and you lose answer visibility; block only the training bot and you do not.
Is llms.txt part of LLM SEO?
It is often sold that way, but none of the OpenAI, Anthropic or Perplexity crawler documentation we read says their search crawlers use llms.txt, and Google says its Search does not use it. Treat it as optional housekeeping, behind crawl access and page content.
Keep reading
How to Rank in ChatGPT: What Gets You Recommended
There is no ranked list in ChatGPT. Learn the three ways a brand reaches an answer, what OpenAI documents, and how to run a fair test of your own visibility.
6 min read
GEO vs SEO vs AEO: What Actually Changes
GEO vs SEO vs AEO compared side by side: what each targets, how it is measured, what the engines document, and which label to use when buying or briefing.
5 min read
What Is llms.txt and Do You Need One?
llms.txt examples, what the proposal says and the evidence: Google does not use it and no assistant crawler page we read says it does. Decide in minutes.
6 min read
Free auditLet us tell you what to fix first.
Send us your website and we will send back a short, honest audit of what is holding it back.
Request a free audit