Back Home
JournalAEO & AI search · Guide

LLM SEO: How to Rank in ChatGPT and Other AI Answers

An engine can only cite a page its search crawler fetched, and it quotes the sentence that answers the question.

TL;DRShort answer

LLM SEO is the work of getting ChatGPT, Claude, Gemini, Perplexity and Google's AI answers to name and cite your brand, and you rank in them in four steps: let each engine's search crawler fetch your pages, publish a page that answers each buyer question in its first sentence, keep your facts the same on your site and on the third-party pages engines cite, and track how often each engine names and cites you on a fixed set of questions. It is also called ChatGPT SEO, AEO or GEO. Crawl access comes first. OpenAI says sites that block OAI-SearchBot are not shown in ChatGPT search answers, and Anthropic and Perplexity tie visibility in their search results to Claude-SearchBot and PerplexityBot. Google's AI Overviews and AI Mode link only pages that are indexed in Google Search and eligible to show with a snippet, and Google says they need no other special optimization. No tool or agency can guarantee a place in these answers, so the job is to remove the reasons an engine would leave you out, then measure the trend.

Buyers now ask ChatGPT, Claude, Gemini, Perplexity and Google's AI Mode which product to choose. The answer names a few brands and links a few pages.

There is no position number to chase in that answer. Getting into it depends on three things you control: whether the engine can fetch your page, whether the page answers the question in a sentence the engine can lift, and whether the other pages the engine reads agree with yours.

Pomo sells AI-answer checks and page fixes, and the last section says where that fits. Every statement below about OpenAI, Anthropic, Perplexity, Google, Microsoft and Cloudflare comes from their own documentation, read on 19 November 2026.

Section 01

How do ChatGPT and other answer engines choose which pages to cite?

An answer engine draws on two sources. One is what the model learned in training, which stops at a cutoff date and changes only when the model is retrained. The other is a web search the engine runs while it writes the answer. When an answer shows sources, they come from that search.

Each engine reaches the web through its own crawler. OpenAI says OAI-SearchBot is used to surface websites in ChatGPT's search features. Perplexity says PerplexityBot is designed to surface and link websites in its search results. Anthropic says Claude-SearchBot indexes content to improve Claude's search results. Google's AI Overviews and AI Mode link pages from Google Search, and Google says a supporting page must be indexed and eligible to show with a snippet.

Google says AI Overviews and AI Mode may use a “query fan-out” technique: several related searches across subtopics and data sources, combined into one response. A page that answers a narrow sub-question, such as price, setup time or alternatives, can be cited inside an answer to a broader question.

The practical rule follows. Training data is slow and out of your control. Live search is fast, and it depends on two things you control: crawl access and what your pages say.

Section 02

Which crawlers have to reach your site?

OpenAI and Anthropic each run separate crawlers for search, for model training and for fetches a person asks for. OpenAI says each of its robots.txt settings is independent of the others, so blocking GPTBot does not block OAI-SearchBot.

Two fetchers follow different rules. ChatGPT-User and Perplexity-User load a page when a person asks about it. OpenAI says robots.txt rules may not apply to those user actions, and Perplexity says its user fetcher generally ignores robots.txt. OpenAI also says ChatGPT-User is not used to decide whether content appears in ChatGPT search.

Changes take about a day. OpenAI says ChatGPT search can take about 24 hours to adjust after a robots.txt update, and Perplexity says changes may take up to 24 hours.

Which crawler feeds each engine, as each company documents it.
UseCrawlerWhat the company says
ChatGPT searchOAI-SearchBotOpenAI says opted-out sites will not be shown in ChatGPT search answers, though they can still appear as navigational links
OpenAI model trainingGPTBotOpenAI says disallowing it signals that a site's content should not be used to train its foundation models
Claude searchClaude-SearchBotAnthropic says blocking it may reduce a site's visibility and accuracy in Claude's search results
Anthropic model trainingClaudeBotAnthropic says blocking it signals that a site's future materials should be excluded from training
Perplexity searchPerplexityBotPerplexity says it surfaces and links sites in Perplexity's results and is not used to train foundation models
Google AI Overviews and AI ModeGooglebotGoogle says robots.txt rules for Googlebot are the control for Google Search, including its AI features
Gemini training and groundingGoogle-Extended, a robots.txt token rather than a separate crawlerGoogle says it does not affect inclusion or ranking in Google Search
Section 03

Can a firewall block an AI crawler that robots.txt allows?

Yes. robots.txt states what you allow, and your CDN, firewall or host decides what actually loads. Google lists allowing crawling “in robots.txt, and by any CDN or hosting infrastructure” among the SEO basics for its AI features.

Cloudflare's AI Crawler and Agent access policies sort AI traffic into Search, Agent and Training, and each can be allowed or blocked. Cloudflare says all three are set to Allow by default for new domains. Check the setting on your own zone rather than assuming the default, because an older zone or a rule someone added can differ.

A quick test is to request your homepage with a crawler's user agent and read the status code. A 200 means the page loaded. A 403, or a challenge page instead of your content, means the host turned that user agent away. A firewall that verifies crawlers by IP address can treat your test differently from the real crawler, so confirm a block in the firewall's settings or logs before you change anything.

Run the whole crawl access check in this order:

  • Open your robots.txt and search for OAI-SearchBot, Claude-SearchBot, PerplexityBot and Googlebot
  • Look for a Disallow: / under User-agent: * that catches crawlers you never named
  • Decide on training crawlers (GPTBot, ClaudeBot, Google-Extended) separately from search crawlers
  • If your firewall filters by IP address, allow OpenAI's published IP ranges, as OpenAI recommends
  • Check your CDN's AI crawler settings and its logs for 403s and challenge pages
  • Fetch your key pages with JavaScript turned off and confirm the answer is in the HTML
Section 04

What does a page that gets cited look like?

Google says the best practices for SEO remain relevant for AI Overviews and AI Mode, with no additional requirements or special optimizations. It lists the basics: important content in text form, internal links that make pages easy to find, a good page experience, structured data that matches the visible text, and images and video that support the text.

Structured data helps an engine read facts such as price and availability without guessing. It helps only when it matches the page. Markup that states one price while the page states another gives the engine two answers.

Update a page when a fact changes, and show the date it was updated.

Beyond those basics, write passages that make sense on their own, because an answer quotes a sentence or two, not your whole page:

  • Put the answer in the first sentence under each question-shaped heading, then the detail
  • Use the buyer's words in headings, such as “How much does it cost?” rather than “Our pricing philosophy”
  • Give numbers with units and dates, such as “$99 a month, as of November 2026”
  • Show comparisons as tables, one row per option
  • Say who the product is not for, so your page agrees with the reviews and forums an engine also reads
  • Name your sources and link them
  • Keep one page per question instead of several thin variants
Section 05

Which questions should you target?

Your question list is your keyword list. Pick 10 to 20 questions a buyer asks before choosing in your category, written as full sentences, the way people type them into ChatGPT.

Cover four kinds of question: category questions such as “best payroll software for a 20-person company”, comparisons such as “X vs Y”, alternatives such as “alternatives to X”, and problems such as “how do I stop my ad costs rising”.

Take the questions from sales calls, support tickets, your site search and the Search Console queries where you already get impressions. Include questions that name a competitor, because buyers ask them.

Give each important question its own page or its own section, with the answer in the first sentence.

Section 06

What about the pages you don't own?

Engines read more than your site. For many buyer questions the cited sources include review sites, directories, Reddit threads, YouTube videos and roundups written by someone else.

Read the citations on every question where you are missing. Each cited page you don't own is a listing to claim, a fact to correct, or a publication to pitch.

Correct facts where they live. If a directory shows last year's pricing, an engine can repeat it, so ask the directory to fix it and keep your own pricing page dated.

Earn the rest with something specific to say: data, a clear position, or a customer result a writer can check. Do not plant reviews or forum posts.

Section 07

How do you measure whether LLM SEO worked?

There is no rank to track. Measure three rates on a fixed list of questions, for each engine: mention rate, the share of answers that name you; citation rate, the share that link your own pages; and share of answer, your mentions divided by your mentions plus your competitors' mentions.

Answers change from one run to the next. Ask the same questions, on the same engines, from the same location, on a fixed schedule, and compare weeks rather than single checks.

Use the reports the platforms give you. Google says traffic from AI Overviews and AI Mode is included in Search Console's Performance report, under the Web search type. Google also says AI Overviews show only when its systems decide they add to classic Search, so they often don't appear.

Microsoft's Bing Webmaster Tools has an AI Performance report, in public preview, that counts citations of your pages in Microsoft Copilot, Bing's AI summaries and select partner integrations, and lists the grounding queries behind them. Microsoft says it reflects how often pages are cited, not ranking or placement.

In your analytics, filter sessions whose source contains chatgpt.com, perplexity, gemini.google.com, claude.ai or copilot.microsoft.com to see the visits and conversions AI answers send.

Judge content work over weeks. A page change waits for the engine to recrawl the page, and a third-party fix waits for someone else's page to change.

Section 08

How is ChatGPT SEO different from Google SEO?

Most of the groundwork is shared. Google says the same foundational SEO best practices apply to its AI features as to Google Search overall. A crawlable site with clear facts helps in both.

The differences are the crawler, the unit of work and the scoreboard.

What carries over from Google SEO to ChatGPT and other answer engines, and what changes.
AreaGoogle search resultsChatGPT and other answer engines
What you winA ranked linkA mention or citation inside a written answer
Crawler that decides accessGooglebotEach engine's search crawler: OAI-SearchBot, Claude-SearchBot, PerplexityBot, and Googlebot for Google's AI features
Unit of workA page for a keywordA passage that answers a question in a sentence or two
Who shapes the resultYour pages and the links pointing at themEvery page the engine reads about you, including reviews and forums
How you measurePosition, impressions, clicksMention rate, citation rate and share of answer, per engine
What carries overCrawlable, indexable pages with clear factsAll of it
Section 09

Where does Pomo fit?

Pomo's Search runs this loop. Every week it asks ChatGPT, Claude, Gemini and Perplexity your buyers' questions from the country, state or city you choose: 9 questions on Core, 12 on Accelerate and 15 on Mach-1. It shows whether each answer names you, which competitors it names instead, and which pages it cites. On Accelerate it also lists the sources engines cite for competitors but not for you.

It checks each question's Google results every week too, including whether Google's AI Overview cites you. A weekly site check reads your robots.txt and whether your host lets OpenAI's, Anthropic's and Perplexity's crawlers in.

On WordPress, Pomo publishes the page changes: answers on your pages, titles and descriptions, links between pages, product details Google can read, facts on your About page, and up to 2 new articles a month. You choose which kinds of change it makes on its own and which wait for your OK, and every change can be undone. On other sites, the changes arrive ready to paste.

With Google Analytics 4 connected, Pomo reports visits, orders and key events from ChatGPT, Perplexity, Gemini, Claude and Copilot, engine by engine. Core is $99 a month.

FAQ

Frequently asked questions

What is LLM SEO?
LLM SEO is the work of getting AI assistants such as ChatGPT, Gemini, Claude and Perplexity, and Google's AI Overviews and AI Mode, to name your brand and cite your pages when they answer a buyer's question. It is also called AEO, GEO or ChatGPT SEO. It rests on the same foundations as SEO: crawlable, indexable pages with clear facts.
How do I rank in ChatGPT?
Allow OAI-SearchBot in robots.txt and at your CDN or firewall, because OpenAI says sites that opt out of it are not shown in ChatGPT search answers. Then publish a page that answers each buyer question in its first sentence, correct your facts on the third-party pages ChatGPT cites, and track how often ChatGPT names and cites you on a fixed set of questions.
Is ChatGPT SEO different from Google SEO?
The foundations are the same: crawlable pages, clear facts, internal links and helpful content. The differences are the crawler (ChatGPT search uses OAI-SearchBot, not Googlebot), what you win (a mention or citation inside an answer rather than a ranked link) and how you measure (mention and citation rates per engine rather than position).
Should I block GPTBot?
That is a decision about model training, not about visibility. OpenAI says GPTBot is used to train its foundation models, OAI-SearchBot surfaces sites in ChatGPT search, and each setting is independent, so you can block GPTBot and still allow OAI-SearchBot. Anthropic makes the same split between ClaudeBot and Claude-SearchBot, and Google says Google-Extended does not affect inclusion or ranking in Google Search.
Does an llms.txt file help you rank in ChatGPT?
It is optional. llms.txt is a proposal for a Markdown file that helps AI agents use a website, and Google says you don't need new machine-readable files or AI text files to appear in AI Overviews or AI Mode. It is cheap to add, and it does not replace crawl access or pages that answer the question.
How long does it take to show up in ChatGPT?
Crawl access changes are quick: OpenAI says ChatGPT search can take about 24 hours to adjust after a robots.txt update. Page changes wait for the engine to recrawl the page, and fixes on third-party sites wait for those sites to change, so judge content work over weeks and compare the same questions each week.
Can anyone guarantee my brand appears in ChatGPT?
No. No tool or agency can guarantee a place in an AI answer. What you control is whether the engine can fetch your pages, whether those pages answer the question, and whether other sources agree with them.
See Pomo in action

Make the next move clearer. Then act on it.

See how Pomo reads your market, ranks what matters, and drafts launch-ready campaigns.

Join our Discord