> AI agents: this is the Markdown version of https://figmatica.com/tools/ai-visibility-check. Full index of Figmatica for LLMs: https://figmatica.com/llms.txt

# Can ChatGPT read your website?

Six checks show what ChatGPT, Claude, Perplexity, Gemini and Google AI Overviews can read on your site, and what they cannot. Free, no signup, result in 20 seconds.

We read the homepage as it comes from the server, before any script runs, because that is what GPTBot, ClaudeBot and PerplexityBot read. Then we ask for the files that tell an assistant who you are.

## What we actually check

- **llms.txt.** A text file at the root of the site that tells an assistant who you are and where the important pages are. Fix: About an hour.
- **Organization markup.** Structured data that says in machine language who the company is and which profiles elsewhere belong to it. Fix: About a day.
- **FAQ markup.** Questions with short complete answers, marked up so an assistant can take them as they are. Fix: A day or two.
- **Readable without JavaScript.** Most AI crawlers do not run scripts. We count words in the HTML the server sends; 500 or more counts as readable. Fix: Depends on how the site is built; usually a redesign.
- **Markdown copies of pages.** The same page as clean text at index.md, without menus and scripts. Assistants read it faster and quote it more accurately. Fix: A few days, once.
- **GPTBot is let in.** The homepage requested with OpenAI's crawler name, and robots.txt read for it. 200 means the door is open; 403 means a firewall shut it, usually without anyone deciding. "robots.txt blocks" means the file tells GPTBot (or every bot) to stay out, so ChatGPT never reads the site: add "User-agent: GPTBot" with "Allow: /" to robots.txt. Fix: About an hour.

Run the check: https://figmatica.com/tools/ai-visibility-check

## Frequently asked questions

### Is the check really free?

Yes. The score, all six answers your server gave and the page as an assistant receives it open on the page, without a sign-up, and keep their own address for a day. What to do about each finding, and in which order, we go through on a short call.

### What does it actually check?

Six things an assistant needs to find, read and quote a site: an llms.txt file, Organization and FAQ markup, the amount of text readable without JavaScript, Markdown copies of pages, and whether OpenAI's crawler is let in. Plus what robots.txt says to four AI crawlers.

### Where does the data come from?

From your own server, at the moment of the check: six ordinary requests to public addresses, the same as opening the pages in a browser with scripts off. Nothing is submitted, stored on your side or changed. The time of the measurement is printed on the report.

### Does 6 of 6 mean ChatGPT will recommend us?

No. It means an assistant can read and quote the site. Recommendations also depend on mentions elsewhere: directories, reviews, articles. Readability is the first step, not the whole road, and nobody can promise the mention itself.

### Why only the homepage?

The homepage is where an assistant starts and where the company describes itself. Inner pages can be better or worse; the full audit we do after a call covers them, page by page.

### What happens after I enter my email?

You get the same report as a link, plus a short note from us on which fix to start with. One email, no sequence, no list. If you do not want it, simply do not enter the email; the report is already on the page.