Free tool. Enter a page address and we show you how much of it an AI like ChatGPT can actually read. It checks technical readability,
not whether AI will cite you.
Enter a page address, for example your own homepage. The check takes a few seconds.
One address, without running JavaScript. It is a snapshot of the moment you run it, and changes when the site changes.
At the end of the check your page gets one of these three answers. The report then tells you exactly which mechanism is stopping it: a noindex tag and a crawler disallowed in robots.txt have different effects, and are fixed in different ways.
Access limited
One or more of the systems we checked cannot reach the page, or the page asks not to be shown in snippets. The effect differs by system: blocking OpenAI's training crawler is not the same as blocking the one that feeds ChatGPT's answers.
What to do: Check in the report which system is blocked, then decide whether that block is intentional. It is often a single line in robots.txt or one tag on the page.
Not readable
Access works, but the page's text is not in the code the server returns: it only appears after the browser runs the JavaScript. Most AI systems do not run it, so they find an empty page.
What to do: Check whether the site can generate the text server-side or with prerendering. The other checks count for little until the text is there.
Readable
The text is in the code the server returns and nothing blocks access. That means this check found no technical obstacle: it is not a guarantee of being cited, which depends on factors this tool does not measure. The remaining results are areas to improve.
What to do: Nothing to unblock. From here the work is on content and on how people find you, not on the technical side.
The tool retraces the four steps an AI system takes when it lands on your page.
Access rules
Reads the site's robots.txt file, the list of rules telling automated systems who may enter, and checks whether it blocks one of the known AI crawlers on this specific address.
Page permissions
Checks whether the page explicitly asks not to be indexed or shown in snippets, via the meta robots tags and the X-Robots-Tag header. These are the signals Google uses to decide eligibility for AI Overviews.
Text in the HTML
Downloads the page's code exactly as the server returns it, without running JavaScript, and measures how much text it actually contains. This is the closest approximation to what a non-rendering crawler receives.
Identity signals
Checks the structured data, the canonical tag, the sitemap and the other signals that help an engine understand what the page is about and who is behind it.
Imagine a site built with a JavaScript framework, with no server-side rendering. Here's what this tool would see, and it's the same thing an AI crawler would see:
| Raw HTML returned by the server | <div id="root"></div>, no text |
|---|---|
| Words counted in the raw HTML | 0 |
| Linked JavaScript | about 850 KB |
| Verdict | Not readable |
In a visitor's browser, the page looks normal: React or Vue build it after the page loads. But GPTBot, ClaudeBot and most AI crawlers don't execute JavaScript: they only receive that empty div.
Here is how the report would show it: the site clears the first step and stops at the second.
Updated 20 August 2026
Most AI crawlers do not run JavaScript. There are two exceptions: Googlebot, which AI Overviews and AI Mode also run through, and Applebot, which uses a real rendering engine.
| GPTBot, OAI-SearchBot, ChatGPT-User (OpenAI) | No |
|---|---|
| ClaudeBot, Claude-SearchBot, Claude-User (Anthropic) | No |
| PerplexityBot, Perplexity-User (Perplexity) | No |
| Meta-ExternalAgent (Meta AI) | No |
| Bytespider (ByteDance) | No |
| Applebot (Apple, Siri and Spotlight) | Yes, via a browser-based crawler |
| Googlebot (Google Search, AI Overviews, AI Mode) | Yes, through Googlebot's rendering service |
Source: Vercel, "The rise of the AI crawler" (GPTBot fetches JavaScript in 11.50% of requests, ClaudeBot in 23.84%, never executing it; Applebot and Gemini do execute it); Lantern, analysis of over 500 million GPTBot requests, June 2026.
The limits of this check
This is a technical check of a single address, on the code the server returns. It does not run JavaScript, it does not impersonate an AI crawler, and it does not measure authority, content quality, or whether an AI engine actually cites the site. It is a snapshot of the moment it runs: the result changes if the site changes.
We repair crawler access, missing raw-HTML text, indexing, canonical and
structured-data issues. And we build sites that are readable
from the server in the first place.