Does this test come from the bot's IP address or just its user-agent?

Only the user-agent header. Requests are sent through this site's own server, not from the real bot IP ranges of the company in question. A security system that verifies IPs may treat this request as an impersonator and block it; the result therefore shows a user-agent-based block, not a definitive verdict.

robots.txt says "allowed" but the test shows "blocked" — how is that possible?

This is exactly the situation the tool exists to measure: robots.txt is a file, and your server can run perfectly well without ever reading it. What actually determines access is the security layer in front of your server (a WAF, a CDN rule, a bot-management product); that layer can operate independently of robots.txt and turn the same bot away with a 403.

Is the address I enter stored anywhere?

No, not permanently. To run the test, our server makes a one-off request to that address; we do not keep the address or the result in a database.

Which user-agents are tested?

A real browser reference, GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot and PerplexityBot, plus Googlebot; the robots.txt analysis additionally evaluates the Google-Extended token (this is a robots.txt rule token, it does not send requests with its own user-agent).

The test showed "allowed" — does that mean I'll be mentioned by ChatGPT now?

No. This tool only tests the access layer (can the bot reach your page). Whether a brand gets mentioned in an answer depends on the model's source selection and the question itself; access is a necessary condition but not a sufficient one on its own.

AI Bot Access Checker: Free GPTBot Test

robots.txt can say "come in" to an AI crawler while the security layer in front of your server (a WAF, a CDN rule, a bot-management product) turns the same crawler away with a 403. The two layers operate independently: one is a file, the other is a runtime decision.

This tool runs both checks together: it downloads your robots.txt and computes which paths are allowed for seven bot tokens (GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, PerplexityBot, Google-Extended, Googlebot); then it sends a real request to the address you give with seven different user-agents and compares the status code, any redirect, and signs of blocking.

Requests are sent from this site's own server (the Cloudflare network), not from the crawler's own infrastructure. What that means and how to read the results is explained further below under "What this tool does not do"; the result is an observation, not a guarantee.

We measured a real example of a security layer returning 403 to GPTBot while robots.txt explicitly allowed access in Can AI Crawlers Reach Your Site? A Verification Guide; this tool automates part of the manual curl test from that guide for you.

How it works

When you enter an address and press "Test", our server first downloads the robots.txt file at the root of that site once (following up to five redirects itself if needed) and computes the Allow/Disallow decision for seven bot tokens — following the group-merging, longest-match, and Allow-wins-on-tie rules set out in RFC 9309.

Then seven separate requests are sent to the address you gave (at most three at a time; the robots.txt request runs separately from these): one carries a real browser user-agent as a reference, the other six carry the strings published in each provider's documentation for GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, PerplexityBot and Googlebot (for a bot whose exact string isn't published, the documented token is used instead). Each request waits at most eight seconds, and redirects (3xx) are not followed — the target address is reported as-is.

The page only shows the status code, duration, content length, the X-Robots-Tag header, a meta robots noindex check, and any redirect address; the actual page content is never stored on our server and is never sent back to you as a body.

Enter an address

Type the full address (starting with https://); you can test your own site or any other address you have permission to test.

What this tool does not do

The result is a point-in-time observation, not a definitive verdict; read the points below before concluding a bot is "definitely blocked" or "definitely allowed".

Requests are sent from this site's own server (the Cloudflare network), not from the real bot IP ranges of the company in question. Systems that do IP verification (some WAF products that check a verified-bot list) may treat the spoofed user-agent as an impersonator and block it — in that case the tool shows "blocked" even though the real bot likely gets through. The reverse is also possible: a system that does not verify IPs may block the real bot while allowing our request.
The result therefore shows a user-agent-based block, not an IP-based decision — it is not a definitive verdict.
The robots.txt analysis, by contrast, is deterministic: the result for the same file is always the same, regardless of network conditions.
A bot successfully reaching a page right now does not mean that bot will mention or rank you in an answer; this tool only tests the access layer.
If the site is under heavy load or returning a temporary error at that moment, the result can be misleading; re-running the test at a different time helps.

Frequently asked questions

Want help tracing the source of a block?

The free business analysis is an initial conversation; we talk through your test result and your server/security layer together.

Request a free business analysis