Launch for free, or pick your own date from $19 one-time. See pricing
Back to Tools

GEO Checker

Check whether ChatGPT, Claude, Perplexity and other AI search engines can fetch and read your site. Every row is a live request to your site, scored out of 100.

What the GEO checker tests

The checker loads your homepage the way a crawler does, then requests robots.txt, llms.txt, llms-full.txt, ai.txt, your sitemap, a markdown version of the page and an MCP discovery file. Points go only to things that change whether an AI engine can fetch and read your site:
  • AI crawler access (25 points): robots.txt lets OAI-SearchBot, ChatGPT-User, Claude-SearchBot, Claude-User, PerplexityBot and Perplexity-User fetch the page.
  • Homepage as crawlers see it (20 points): Status 200, not noindex, and the main text is in the HTML the server sends, not built later by JavaScript.
  • Structured data (15 points): Valid JSON-LD, including an Organization or WebSite entity that says who runs the site.
  • llms.txt (15 points): A plain-text /llms.txt that follows the llmstxt.org format.
  • Page metadata (10 points): Title, meta description and canonical URL.
  • Sitemap (10 points): A valid XML sitemap, found through robots.txt or at /sitemap.xml.
  • Markdown version (5 points): A markdown copy of the page via Accept: text/markdown, /index.md or a rel=alternate link.
AI training crawlers, llms-full.txt, ai.txt and MCP discovery are reported but not scored. Each check is a live request made when you press the button, so the result reflects your site right now, not a cached crawl.

AI crawlers: which ones decide whether you get cited

AI companies run separate crawlers for search, for fetching a page while a user waits, and for collecting training data. Only the first two affect whether an answer can link to you. This is the list the checker tests:
User agentCompanyWhat it doesAffects citations
OAI-SearchBotOpenAIBuilds the index ChatGPT search shows and links toYes
ChatGPT-UserOpenAIFetches a page when a ChatGPT user or GPT asks for itYes
GPTBotOpenAICollects training data for OpenAI modelsNo
Claude-SearchBotAnthropicIndexes pages to improve Claude's search resultsYes
Claude-UserAnthropicFetches a page when a Claude user asks a question that needs itYes
ClaudeBotAnthropicCollects training data for Claude modelsNo
PerplexityBotPerplexityBuilds Perplexity's search indexYes
Perplexity-UserPerplexityFetches a page when a Perplexity user asksYes
Google-ExtendedGoogleNot a crawler: a robots.txt token that controls use in Gemini training and grounding. AI Overviews come from regular GooglebotNo
Applebot-ExtendedAppleA robots.txt token that controls use in Apple's AI trainingNo
CCBotCommon CrawlOpen web archive that many AI companies train onNo
BingbotMicrosoftBing's search index, which Copilot answers are built onYes
If you block nothing, every one of them is allowed. Most sites that fail this check didn't block AI search on purpose: a security plugin, a CDN bot setting or an old "block all AI" snippet did it for them.

A robots.txt that lets AI search in

Keep AI search crawlers under the same rules as everyone else, and opt out of training separately if you want to:
robots.txt
# Every crawler, AI search crawlers included
User-agent: *
Disallow: /admin
Disallow: /dashboard

# Optional: opt out of AI training. Blocking these does not
# stop ChatGPT, Claude or Perplexity search from citing you.
User-agent: GPTBot
User-agent: ClaudeBot
User-agent: Google-Extended
User-agent: Applebot-Extended
User-agent: CCBot
Disallow: /

Sitemap: https://example.com/sitemap.xml
The most common mistake is adding a group for one crawler to say it's welcome. Crawlers obey only the most specific group that names them, so that group throws away every rule in the * group:
Don't do this
User-agent: *
Disallow: /admin
Disallow: /checkout

# Meant as "OpenAI is welcome", but a named group
# REPLACES the * group for that crawler: OAI-SearchBot
# may now crawl /admin and /checkout.
User-agent: OAI-SearchBot
Allow: /
If you do need a named group for a search crawler, copy your Disallow lines into it. The checker reads robots.txt the same way the crawlers do (RFC 9309: most specific group, longest matching rule, Allow wins a tie), so a pass here means the rules work as written.

How to fix the common failures

Almost no text in the HTML. Your page is built in the browser. Render the main content on the server: in Next.js, fetch data in a server component instead of in a useEffect; for a Vite or Create React App site, prerender the public pages or move them to a framework that renders on the server, such as Next.js or Astro. Go Publicly moved from a client-rendered React app to Next.js in 2026 for this reason.
No Organization entity. Add one JSON-LD block to the homepage. It tells search and AI engines who runs the site and links your social profiles to it:
In the homepage <head>
<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "Organization",
  "name": "Acme",
  "url": "https://acme.com",
  "logo": "https://acme.com/logo.png",
  "sameAs": [
    "https://x.com/acme",
    "https://www.linkedin.com/company/acme"
  ]
}
</script>
No llms.txt. Write one in a few minutes with the llms.txt generator, then check it with the validator. If /llms.txt returns your homepage HTML, your app is answering every unknown path with index.html: put the file in the public or static folder so it's served as text.
No sitemap. Most frameworks and CMSs can generate one. List every public page, leave out login, admin and search pages, and add a Sitemap: line to robots.txt.
Title, description or canonical missing. The on-page SEO checker audits a single page in more detail.

Example: go-publicly.com scores 95

We ran the checker on our own site. It passes every scored check except the markdown version (5 points): all six AI search crawlers are allowed, the homepage sends about 770 words of text in its HTML, the JSON-LD includes Organization, WebSite and FAQPage, and llms.txt has 5 sections and about 30 links. We haven't added a markdown version because it only saves tokens for agents that fetch pages live; it doesn't change whether we can be cited.

After the fixes: get mentioned

Answer engines cite pages that other sites talk about. Once crawlers can read your site, list it where people look for new products: launch it on Go Publicly and submit it to startup directories.

GEO checker: common questions

What is GEO?

GEO (generative engine optimization) is making a site easy for AI answer engines such as ChatGPT search, Claude, Perplexity and Google AI Overviews to fetch, understand and cite. The technical part is access and readability; the rest is having content and mentions worth citing.

Does a high score mean ChatGPT will cite my site?

No. The score measures whether AI crawlers can reach and read your pages. Whether an answer cites you depends on your content and on how often other sites mention and link to you. A low score can stop citations; a high score doesn't create them.

Why don't ai.txt and MCP discovery count toward the score?

They're proposals that no major AI company has said it reads. The checker still reports whether you have them, but it only awards points for things that affect whether a crawler can fetch and parse your pages.

Should I block GPTBot and other training crawlers?

That's your choice, and the checker doesn't score it. Training crawlers (GPTBot, ClaudeBot, Google-Extended, CCBot) are separate from the search and user-fetch crawlers that answer engines use to cite pages, so blocking training doesn't stop citations.

Does blocking Google-Extended remove me from AI Overviews?

No. Google-Extended only controls whether Google may use your content for Gemini models. AI Overviews are part of Google Search and use the normal Googlebot crawl, so the only way out of them is to restrict Googlebot itself, which also affects regular search.

Why does server-rendered text matter?

Most AI crawlers read the HTML your server sends and don't run JavaScript. If your page builds its content in the browser, as many single-page apps do, those crawlers see an almost empty page.

Can I share my result?

Yes. After a check, the page address includes your site (for example /tools/geo-checker?url=example.com), and opening that link runs the check again.