Is ChatGPT reading my website? How to check

The question has four answers, because four different OpenAI bots can touch your site and they do different jobs. “Is ChatGPT reading my website” usually means one of two much more specific things: can I show up when someone asks about my industry, or is my content being used to train the model. Those are controlled separately.

The four bots, and which one you mean

• OAI-SearchBot — Surfaces websites in ChatGPT's search features. Follows robots.txt.
• GPTBot — Crawls content that may be used to train OpenAI's foundation models. Follows robots.txt.
• ChatGPT-User — Fetches a page live when a person or a custom GPT asks about it. Does not follow robots.txt.
• OAI-AdsBot — Validates the safety of pages submitted as ads on ChatGPT. Follows robots.txt.

If you want to be in ChatGPT’s answers, the bot that matters is OAI-SearchBot. If you want to be out of the training data, the one that matters is GPTBot. Confusing the two is the single most common self-inflicted wound in this area. The full list, including Anthropic, Perplexity, Google and Apple, is on the AI crawler reference.

Check one: is anything blocked?

Open https://yourdomain.com/robots.txt in a browser and read it. You are looking for any Disallow under the tokens above, and for a blanket User-agent: * with Disallow: /, which blocks everything. A rule with a misspelled token matches nothing at all, so read the strings character by character.

Worth knowing: a page can also be blocked without robots.txt saying so. A noindex meta tag, an X-Robots-Tag header, a login wall, a firewall rule or a bot-protection product can each stop a crawler while your browser sees the page perfectly.

Check two: your server logs. This is the real answer.

Logs are the only place where “did it actually happen” is recorded. Every fetch leaves a line with the user agent, the path and the time. If you have shell access:

grep -iE "OAI-SearchBot|GPTBot|ChatGPT-User|ClaudeBot|Claude-SearchBot|PerplexityBot" access.log \
  | awk '{print $1, $4, $7}' | tail -50

On managed hosting, the same thing lives under a name like “access logs”, “traffic” or “bot analytics”. Cloudflare, Vercel, Netlify and most CMS platforms will show you requests by user agent. No lines for these tokens means no OpenAI bot has been to your site, and no amount of on-page work changes that until it does.

Check three: can it use the page once it arrives?

Being fetched is necessary and not sufficient. A crawler that renders no JavaScript sees whatever your server sent as raw HTML. If your content is assembled in the browser, the bot got an empty shell and left without an error anyone will ever see.

The quick version of this test: view your page’s source, or fetch it with curl, and look for the actual words of your content. Whatever is missing there is invisible to a meaningful share of the machines you are trying to reach. That is what a free readiness scan automates: it fetches your site the way these crawlers do and reports what survived the trip.

Check four: ask it yourself, carefully

Asking ChatGPT about your own website triggers a live fetch by ChatGPT-User, so it proves the page is reachable right now. It proves almost nothing else. You told the model which site to look at, so of course it found it. The question that matters is whether it reaches you when it has not been told — which is why the citation self-test in a scan report gives you a blinded prompt that never names your domain.

In order

1. Read your robots.txt. Unblock what you did not mean to block.
2. Search your logs. Establish whether anything is arriving at all.
3. If nothing is arriving: you have a discovery problem, not a technical one. Links and a submitted sitemap fix that; nothing on the page does.
4. If bots arrive but you are never cited: you have a content problem. Start with why a site does not show up in AI answers.