Can ChatGPT see my website? Four checks before changing anything

ChatGPT may be able to fetch your public website while still leaving it out of an answer. Check the exact page, crawler permissions, indexing controls and an actual answer separately before paying for a visibility fix.

A missing mention does not identify the cause. A site owner might be looking at a private preview, a firewall refusal, an outdated page, or an accessible page that was not selected for that question. Start with one important public URL, such as your main service page, and keep a short record of what each check found.

Can someone reach the same page without signing in?

Open the final HTTPS address in a signed-out browser. Check the page you want cited, including any redirect. A working homepage does not establish that your service, pricing or booking page works.

Record the address, date, response and visible main text. An HTTP 200 response is a useful access observation; check that it contains the intended page rather than a login screen, a security challenge or an empty application shell. If the browser presents a certificate or security warning, stop and ask the host to resolve it.

For a UK service business, choose the page explaining its actual coverage and service. A generic homepage saying that the team offers quality solutions leaves a visitor with less to verify than a page listing the places served, the work offered and how an enquiry is handled.

Which OpenAI crawler should I allow?

OpenAI's crawler documentation identifies OAI-SearchBot for search and GPTBot for training. Their settings are independent. Choose the permissions that match your publishing policy.

Start with the free AI crawler checker. Read its findings for the exact URL. Keep a copy of your existing robots.txt before changing it, and ask your hosting provider to review firewall rules when the request is being refused.

A user-agent probe is a simulated request. It can reveal that a server treats an agent name differently; it does not demonstrate that a genuine OpenAI crawler visited. Corroborate real visits with server logs and the provider's published verification information. Avoid disabling the entire firewall to make a test turn green.

Could the page be reachable but excluded from search?

A page can load normally for a visitor while carrying indexing or preview restrictions. Inspect the page directives and response headers, and have the person responsible for the site confirm that the intended public page is eligible for discovery.

Google's AI-feature guidance says its supporting pages need to be indexed and eligible for a search snippet. This is a Google requirement, rather than proof of how another assistant will handle the same URL.

Use this diagnosis sheet to assign the next action:

These observations describe different parts of the process. Keep them in separate columns so a higher technical score cannot accidentally become a claim that assistants are recommending the business.

Does the page answer the question a customer asks?

Read the opening paragraph as a potential customer. It should explain what the service does and the decision the page helps them make. Add evidence that can be checked: actual service details, a dated price basis where appropriate, original photographs or a clear worked example.

For example, a local repair page can explain the work covered by its call-out charge, the area served and how customers receive a quote. Inventing a lowest price or a customer success story would weaken that evidence. Use the answer-readiness checker as a content checklist, then assess the page yourself.

How do I find out whether the change helped?

Keep the original question and run a fresh observation after making a documented change. Record the assistant, available search mode, exact question and cited URLs. Testing with the answer already in the conversation can influence the result.

Follow our free visibility measurement method for a repeatable log. AEO, SEO and GEO explained sets out which improvements belong to access, content and observed results. A change in one column should lead to a specific next check.

Which questions come up most often?

Does passing a crawler check guarantee a citation?

It establishes the access signals tested by that checker. Record actual assistant citations separately for the questions that matter to your business.

Should I change every crawler rule at once?

Keep the existing file and publishing policy, identify the relevant rule, and make a documented change. Review hosting restrictions separately.

Which page should I test first?

Choose the public page that answers a valuable customer question. Check its final address and main text, then repeat the observation after a change.