Can ChatGPT see my website, and how do I check?
There are two ways to check, and they answer different questions. Asking ChatGPT tells you what it remembers or can find right now. Looking at `robots.txt` and your server responses tells you whether it has any way to read you at all — and without that, the first answer will always be "I don't know".
Start with the second one. If the crawler is not let in, no amount of content work changes anything: the text is simply never read.
Who actually visits
OpenAI runs several different bots, and they do different jobs:
GPTBot collects content for model training. Blocking it keeps your text out of what the model "knows" in advance.
OAI-SearchBot indexes content for ChatGPT search. Blocking it removes you from answers that carry links.
ChatGPT-User arrives when a specific person asks the assistant to open your link. Blocking it means your visitor gets "could not reach the page".
Three different bots mean three separate decisions. A common mistake is
blocking all of them with one Disallow, then wondering why a link sent to a
client will not open in ChatGPT.
Checking the technical side
Open yourdomain.com/robots.txt and look for those names. If you see
Disallow: / under any of them, that bot is not welcome.
The second thing is invisible in robots.txt: your server or CDN can block
a bot by User-Agent regardless of what robots says. Cloudflare has a
separate setting for exactly this, and it is often enabled without anyone
deciding to.
This is a real and common situation: robots.txt says "welcome" while the
CDN answers 403.
Checking what the model knows
Ask directly: "what do you know about [your domain]" and "what services does [your company] provide". Pay attention not to whether the answer sounds polished, but to whether it contains facts that exist nowhere else — prices, a city, a specific offer.
If the answer is generic ("it appears to be a website about..."), the model either never read you, or read text with nothing concrete in it.
What is not enough
Letting the bots in is necessary but not sufficient. Having read a page that says "we provide high quality services", a model has nothing to quote — no fact, no number, no answer to any question.
So the work has two halves: access and content. The first takes five minutes; the second is ongoing.
A free scan checks both: whether bots get through (live, not by
reading robots.txt), whether structured data exists, and whether the pages
contain questions with answers at all.
Frequently asked
Is blocking GPTBot a bad idea?
It depends on the business model. For a publisher whose revenue comes from visits, blocking can be the right call. For a service company it means disappearing from answers that could have recommended it.
Does ChatGPT see my changes immediately?
`ChatGPT-User` does — it opens the page at the moment someone asks. What the model "knows" without a link refreshes far less often: training data is rebuilt rarely.
Do I need separate content for AI assistants?
No. You need the same content written so the answer can be found: question in the heading, answer immediately below, facts with numbers. A separate text "for robots" is an old idea that used to earn sites penalties.
How is ChatGPT different from Google AI Overviews?
In the source. AI Overviews draws on Google's index, so ordinary SEO applies. ChatGPT uses its own crawlers and its own data, so being in Google's index does not automatically make you reachable for it.
Check yours
A free scan shows which of the things described here your site already does, and which it does not.