AI Visibility Checker: Can ChatGPT, Gemini and Perplexity Reach Your Website?
Short answer: An AI visibility checker does one of two jobs. It either tracks how often ChatGPT, Gemini or Perplexity mention your brand, or it checks whether AI crawlers can reach and read your pages at all. Start with the second: no tracking report helps if a crawler is blocked or your text only appears after JavaScript runs. SEOgrinder runs that readiness check for free, without an account.
What an AI visibility checker actually measures
The phrase covers two different products, and mixing them up leads to the wrong fix.
Mention tracking sends a fixed list of questions (prompts) to AI assistants and records whether your brand or your address shows up in the answers. It answers the question "Are we mentioned?". Free and paid trackers exist. Ahrefs, for example, offers a free AI visibility checker that reports how often AI platforms mention and cite a brand.
Readiness checking looks at your own website. It answers "Can an AI system fetch and read our pages?". That is a technical question, and most checks have a clear yes or no: is the crawler allowed in robots.txt, does the delivered HTML contain the text, is the page indexable. A readiness check does not tell you whether an assistant will mention you. It tells you whether you have removed the obvious blockers.
Both are useful. They answer different questions, and the order matters.
Why readiness comes first
Three documented facts explain it:
- ChatGPT search relies on OAI-SearchBot. OpenAI states that sites opted out of OAI-SearchBot are not shown in ChatGPT search answers, although they can still appear as navigational links. A change in robots.txt can take about 24 hours to take effect. Source: OpenAI crawler documentation.
- AI crawlers mostly read the HTML they are given. A December 2024 analysis by Vercel and MERJ found that none of the major AI crawlers it studied executed JavaScript, while Googlebot (for Gemini) and AppleBot did. It has not been repeated in the sources we checked, so read it as a strong warning, not as a permanent law. Source: Vercel, The rise of the AI crawler.
- Google's AI features have no extra requirements. Google says there are no additional requirements to appear in AI Overviews or AI Mode, and no special optimizations. A page has to be indexed and eligible to appear in Google Search with a snippet. Source: Google Search Central.
Put together: if the door is shut, nobody can read the room. A tracker will show you a low number, but it will not tell you that your robots.txt blocks the bot or that your page is an empty shell until a script runs.
The checks that decide whether AI can read your site
| Check | Question it answers | Typical failure |
|---|---|---|
| Rules for AI crawlers in robots.txt | Is each bot allowed or blocked on purpose? | One broad Disallow: /, or a bot group that forgot the shared rules |
| Search bot versus training bot | Do you block training but still allow search? | Blocking GPTBot and believing ChatGPT search is off |
| Text in the delivered HTML | Is the main content in the first response, without JavaScript? | An empty page shell with a script tag |
| Indexable with a snippet | Is the page free of noindex and nosnippet? | A staging setting that went live |
| Access through CDN and firewall | Do bots get a normal page, not a block? | A bot-protection rule that rejects unknown crawlers |
| Sitemap | Are the pages you want found listed? | No sitemap, or a sitemap full of redirects |
| Optional signals | Content-Signal, llms.txt, Markdown version | Treated as ranking levers, effect not proven |
Google names the first, fourth and fifth row directly. A page must be indexed and eligible to appear in Search with a snippet, and as a best practice crawling should be allowed in robots.txt and by any CDN or hosting infrastructure. Google also lists internal links and important content in text form among the basics.
Which AI crawler does what
OpenAI and Anthropic separate training, search and user requests; Perplexity and Google divide the work differently. That is why one User-agent line is rarely enough.
| Bot | Operator | Used for | If you block it |
|---|---|---|---|
| GPTBot | OpenAI | Training of models | Signals that content should not be used for training |
| OAI-SearchBot | OpenAI | Search in ChatGPT | The site is not shown in ChatGPT search answers |
| ChatGPT-User | OpenAI | A user's request in ChatGPT | robots.txt rules may not apply, because a person asked |
| ClaudeBot | Anthropic | Collecting content for training | Signals that future content should be excluded from training |
| Claude-SearchBot | Anthropic | Search result quality | Visibility in search results may drop |
| Claude-User | Anthropic | A user's question to Claude | Retrieval on request is prevented, visibility may drop |
| PerplexityBot | Perplexity | Search results in Perplexity | Not used for training; Perplexity recommends allowing it |
| Google-Extended | Gemini training and grounding | No effect on inclusion or ranking in Google Search |
Sources: OpenAI, Anthropic, Perplexity and Google. Operators change their lists, so check the pages before you copy a rule.
Run the readiness check in five minutes
- Open SEOgrinder, type your address and start the analysis. No account is needed.
- Open the AI readiness tab in the report.
- Read "Checks for AI agents". It has three basics (robots.txt, sitemap, rules for AI crawlers) and four extras (Content-Signals, llms.txt, Markdown version, link headers). The extras are suggestions from individual vendors, and while one of them is open, the tool says so: an effect on Google is not proven.
- Met points can be expanded: they show the text that was found in your files. Note that SEOgrinder fetches your site under its own name, not as GPTBot.
- Press Copy agent prompt and paste it into your coding assistant. For an open point the prompt describes what to create and how to check it; for a met point it asks the assistant to review what exists and change only what is wrong.
- Press Check again to confirm. Reports are cached for up to 24 hours, so use that button after a fix; it works a few times per hour.
The report also flags pages that only build their interface in the browser. For AI crawlers that do not run JavaScript, such a page can look empty.
Add tracking afterwards, and start by hand
Once the basics are green, measure the outcome. You do not need a paid tool to begin:
- Write down ten questions a customer would really ask. For a local ice cream shop, that is "best ice cream in [your city]" and a few variations.
- Ask each question in ChatGPT, Gemini and Perplexity, in a fresh chat.
- Log the date, the assistant, the question, whether you were mentioned, which pages were cited and who else appeared.
- Repeat monthly with the same questions.
Answers differ from run to run and from user to user, so watch the trend over several rounds instead of reacting to one result. If you later need volume, a paid tracker can run the same prompts automatically.
Common mistakes
- Blocking everything after a scare.
Disallow: /for all bots also removes you from search results. - Blocking GPTBot to leave ChatGPT. GPTBot is about training. ChatGPT search uses OAI-SearchBot, and each setting is independent.
- A separate group that drops the shared rules. A crawler that finds its own group follows that group and ignores the
*group (RFC 9309). List severalUser-agentlines above one set of rules, or repeat the rules. - Reading a tracker score as proof of readability. A page that is empty without JavaScript can still be mentioned from other sources, and a readable page can still be ignored.
- Treating llms.txt as a ranking lever. Google's Gary Illyes is reported to have said that Google does not support llms.txt, and John Mueller said that, as far as he knew, no AI system used it. These are second-hand reports, see Search Engine Roundtable. The file is cheap to add but not proven.
- Expecting an instant change. Give crawlers about a day after a robots.txt update.
FAQ
Is there a free AI visibility checker?
Yes, of both kinds. Mention checks such as the one from Ahrefs show whether AI platforms cite a brand. SEOgrinder checks whether AI systems can reach and read your site, for free and without an account.
Can a checker guarantee that ChatGPT mentions my business?
No. Nobody controls what an assistant answers. You can remove technical blockers and publish pages that are worth citing. Whether an assistant names you is its decision.
Does blocking GPTBot take me out of ChatGPT answers?
No. GPTBot is the training crawler. ChatGPT search can only show sites that allow OAI-SearchBot, and user requests use ChatGPT-User.
Do I need an llms.txt file?
It is optional. It costs little, and SEOgrinder checks for it, but a ranking or citation effect is not proven. It never replaces robots.txt, a sitemap or pages that answer real questions.
Which check should I do first?
The readiness check. If a crawler cannot fetch your page, every later measurement starts from zero.
Your next step
Run the free scan on your own site and open the AI readiness tab. If one of the three basics is open, fix that first, press Check again, and only then start tracking mentions.