LLM SEO: a practical guide to getting found and cited by ChatGPT, Claude, Gemini, and Perplexity

This guide is for SaaS marketers, SEO leads, and developers who want their site to show up when people ask ChatGPT, Claude, Gemini, or Perplexity about their category. It is deliberately practical: what each engine has documented about how it finds pages, what to change on your site this week, and how to tell whether it worked.

Every technical claim here comes from the vendors’ own documentation, fetched on October 29, 2026. Crawler names and rules change, so check the linked pages before you edit robots.txt. Where a vendor has not documented something, we say so.

Page ownership: this URL owns LLM SEO as a hands-on, engine-by-engine how-to. The full generative engine optimization system lives on /blog/generative-engine-optimization. What LLM visibility means and how to track it lives on /blog/llm-visibility. The SaaS implementation cadence lives on /blog/ai-search-optimization-playbook. This page links to those instead of repeating them.

LLM SEO, also called LLM optimization, LLMO, or LLM search optimization, is the practice of making a website easy for large language model products to find, read, trust, and cite. Most AI assistants that answer questions about current products do not rely only on what the model memorized in training. They search the web, retrieve pages, and write an answer grounded in what they found, often with source links. LLM SEO targets that retrieval step.

Google describes the same mechanism for its own AI features. Its generative AI guide says AI Overviews and AI Mode are rooted in core Search ranking and quality systems, and use retrieval-augmented generation (grounding) plus query fan-out, a set of related searches the model issues to gather more results. Google’s position is that optimizing for its AI features is still SEO. That is a useful anchor: LLM SEO is not a separate discipline with secret rules. It is search optimization with a few new crawlers and a stronger premium on clear, quotable answers.

Each vendor separates crawling for search from crawling for training. That split is the single most useful fact in LLM SEO, because it lets you decide independently whether your content can be cited in answers and whether it can be used to train models.

ChatGPT. OpenAI says OAI-SearchBot is used to surface websites in ChatGPT’s search features, and that sites opted out of OAI-SearchBot will not be shown in ChatGPT search answers, though they can still appear as navigational links. GPTBot crawls content that may be used for training, and OpenAI says the two settings are independent. ChatGPT-User fetches pages when a user’s request needs them; OpenAI says robots.txt rules may not apply to it because the action is user-initiated, and it is not used to decide search inclusion. OpenAI says robots.txt changes take about 24 hours to affect search.

Claude. Anthropic’s help article (April 7, 2026) lists three bots. ClaudeBot collects content that may contribute to training. Claude-User retrieves pages when a user asks Claude a question; Anthropic says disabling it may reduce visibility for user-directed web search. Claude-SearchBot indexes content to improve search results; disabling it may reduce visibility and accuracy in search results. Anthropic says its bots honor robots.txt and the Crawl-delay extension.

Perplexity. Perplexity’s docs say PerplexityBot surfaces and links websites in Perplexity search results and is not used to crawl content for AI foundation models. Perplexity-User supports user actions and, because the fetch is user-requested, generally ignores robots.txt. Perplexity says changes can take up to 24 hours and publishes WAF allow-list guidance.

Gemini and Google Search. For AI Overviews and AI Mode, Google says robots.txt rules for Googlebot are the control, and a page must be indexed and eligible to show with a snippet. There are no additional technical requirements. For the Gemini app, Google-Extended matters: Google’s crawler docs describe it as a robots.txt token (not a separate crawler) that controls whether crawled content may be used to train Gemini models and to ground answers in Gemini Apps and Vertex AI. Google says it does not affect inclusion or ranking in Google Search.

If a search crawler cannot reach a page, no amount of content work will get that page cited by that engine. Start with an AI crawler robots.txt audit. Check robots.txt for each user agent above, then check what actually happens at the edge: CDN bot protection, firewall rules, and rate limits can block a bot that robots.txt allows. Google’s AI features page lists “crawling is allowed in robots.txt, and by any CDN or hosting infrastructure” as its first best practice, and both OpenAI and Perplexity recommend allowing requests from their published IP ranges.

Decide on training separately. Many teams allow the search crawlers so they can be cited, and make a deliberate, documented choice about GPTBot, ClaudeBot, and Google-Extended. Each vendor says these settings are independent, so a training opt-out does not need to cost you search visibility.

Make sure the content is actually in the HTML the crawler receives. Google says it can process JavaScript that is not blocked, but other crawlers document less about rendering. Server-rendered or static HTML for key facts, prices, and answers is the safe default.

Once a crawler can reach the page, the question becomes whether the page contains something worth citing. Google’s guide is direct about what helps most in the long run: unique, non-commodity content with a real point of view, first-hand experience, and clear organization by sections and headings. A summary of what everyone else already says gives an AI system nothing to add.

Put a direct answer near the top of each page, in two or three sentences that make sense on their own. Repeat the subject instead of starting with “It.” Follow with the evidence: numbers you can source, dates, steps, and comparisons. Mark facts that change with a date, because stale pricing and feature claims are a common source of wrong AI descriptions.

Cover the questions a buyer asks next. Google says AI Overviews and AI Mode use query fan-out, issuing related searches across subtopics. A page that answers the main question and its obvious follow-ups is more likely to be retrieved for one of them. But Google also warns that creating separate pages for every fan-out variation, primarily to manipulate rankings or AI responses, violates its scaled content abuse policy. Answer follow-ups in depth on the page that owns the topic.

LLMs combine what several sources say. If your site, your profiles, your directory listings, and independent coverage describe your product in different terms, the assistant has to guess. Pick one plain description of what you are, who it is for, and what it costs, and use it everywhere you control.

Then earn corroboration the honest way. Google’s guide notes its AI features can show what is said about products across blogs, videos, and forums, and warns that seeking inauthentic mentions is not as helpful as it might seem. Real reviews, real comparisons, and real discussions carry more weight than placed mentions. The full method for entities and corroboration is on /blog/generative-engine-optimization and /blog/entity-seo-for-saas.

The fundamentals are shared, but each engine has its own documented controls and quirks. If you need to optimize for ChatGPT first because that is where your buyers ask, start with its row, then work across. Use this as a starting point, then go to the engine guide for depth.

There is no LLM ranking report to download. Measure in three layers. First, crawl: server logs or a bot analytics tool show whether OAI-SearchBot, Claude-SearchBot, PerplexityBot, and Googlebot are fetching your pages. Second, referrals: analytics shows visits from chatgpt.com, perplexity.ai, claude.ai, and gemini.google.com. Third, answers: a dated prompt log records whether assistants name and cite you for the questions your buyers ask.

For Google, use Search Console. Google’s AI features page says appearances in AI Overviews and AI Mode are counted in the Performance report under the Web search type, and its generative AI guide points to a generative AI performance report as well. Google also warns that no third-party tool has access to its internal ranking or AI systems.

Here is what this looks like on TopFound’s own small site, labeled and dated. In the 30 days to October 27, 2026, DataFast logged OpenAI as the most active indexing bot on topfound.io (about 1,400 hits), ahead of Perplexity (577), Bing (398), and Google (311). In Google Search Console for September 30 to October 27, 2026, our /blog/google-ai-mode post earned 1,950 impressions for “google ai mode” at an average position of 11.6, while /blog/ai-search-analytics (average position 30.4) and /blog/ai-citation-optimization (33.8) had the next-best positions among our top pages. These are our own readouts, not benchmarks for anyone else. The full measurement method is on /blog/ai-search-analytics.

Teams can run this guide themselves. Teams that want TopFound to audit crawler access, rewrite key pages, and run a dated prompt log across ChatGPT, Claude, Gemini, and Perplexity can start with a free AEO report at /aeo-report or the LLM visibility service at /services/llm-visibility-optimization. Public plans are on /pricing: Growth $797/month, Scale $1,297/month, and Custom scoped.

LLM SEO comes down to four steps. Let each engine’s search crawler in, at robots.txt and at the edge. Write pages with direct, dated, self-contained answers and real evidence. Describe your brand the same way everywhere and earn genuine corroboration. Measure crawl, referral, and answer data separately, without pretending there is a single AI rank.

None of that needs a trick, and the vendors’ own documentation says as much. For the full system, continue to /blog/generative-engine-optimization; for tracking, /blog/llm-visibility; for paid placement in ChatGPT, /blog/chatgpt-ads.