Blocking the Wrong AI Bot Could Hide You From ChatGPT

You can block OpenAI from training on your website and still let your business appear in ChatGPT search. You can also do the opposite by accident.

That distinction matters if you run a Perth trade or service business and want to be recommended when someone asks ChatGPT for a provider nearby. OpenAI uses separate controls for search, model training and user-requested page visits. A blanket rule copied into robots.txt can protect the wrong thing while removing the visibility you wanted.

This guide explains what each control does, where the common advice goes wrong and what to check before anyone changes your site.

One company, three AI bots, three different jobs

OpenAI’s crawler documentation separates search visibility, model training and user-requested visits. The settings are independent. That means you can allow search access while declining training access, rather than treating every OpenAI request as the same thing.

OAI-SearchBot controls ChatGPT search visibility

OAI-SearchBot is the search control. OpenAI says sites that opt out “will not be shown in ChatGPT search answers, though can still appear as navigational links”. If being found in ChatGPT search matters to your business, this is the token you do not want blocked by accident.

GPTBot controls model training

GPTBot crawls content that may be used to train OpenAI’s generative AI models. Disallowing it tells OpenAI not to use your site’s crawled content for that purpose. It does not opt your site out of ChatGPT search. Search is controlled separately by OAI-SearchBot.

ChatGPT-User handles visits requested by people

ChatGPT-User may visit a page when somebody asks ChatGPT or a custom GPT to fetch it. OpenAI says these actions are initiated by a user, so robots.txt rules may not apply. It also says this agent does not determine whether content can appear in search.

Why “block the ChatGPT bot” is bad advice

There is no single ChatGPT bot switch. Advice that treats OpenAI’s agents as one crawler leaves out the decision that matters: what are you trying to prevent, and what visibility are you prepared to give up?

A WordPress security plugin or copied template may include a blanket “block AI” rule. That sounds tidy. It can also block OAI-SearchBot, removing the site from ChatGPT search answers, while failing to guarantee that a user-requested visit through ChatGPT-User will stop.

For a local electrician, plumber or professional service firm, the commercial cost is straightforward. The business may disappear from one place customers are starting to research providers, without getting the hard access block the owner thought they had bought.

If you are still deciding how much AI search matters to your market, our guides to how AI search affects trade businesses and how to prepare a website for AI search cover the wider picture.

A Perth electrician’s robots.txt mistake

Imagine a Perth electrician running a standard WordPress site. The owner reads that AI companies are taking website content and enables a plugin option labelled “block AI crawlers”. The plugin adds this:

User-agent: OAI-SearchBot
Disallow: /

User-agent: GPTBot
Disallow: /

User-agent: ChatGPT-User
Disallow: /

The three rules look consistent. Their effects are not.

• The OAI-SearchBot rule opts the electrician out of ChatGPT search answers.
• The GPTBot rule tells OpenAI not to use the site’s crawled content for training.
• The ChatGPT-User rule is not a guaranteed block because the request may be initiated by a person.

The owner has traded away search visibility but has not built a reliable access barrier. That is a poor exchange, especially when it happened through one vague checkbox.

A corrected version might keep search access open while making a separate decision to decline training access:

User-agent: OAI-SearchBot
Disallow:

User-agent: GPTBot
Disallow: /

This is an example, not a blanket recommendation. Another business may choose to allow both agents. The important part is that the decision is deliberate and token-specific. Because OpenAI says ChatGPT-User rules may not apply, a genuine hard block should happen at the server or CDN layer using verified request data, not through wishful thinking in a text file.

Robots.txt is a request, not a lock

A robots.txt file tells compliant automated crawlers which areas they should access. It is not an authentication system, firewall or legal force field. If access must be prevented, the site needs an enforceable server or CDN rule.

Server and CDN logs also show what is actually reaching the website. User-agent labels can be copied, so a proper audit should compare requests with the IP ranges published by OpenAI and check what the site served. The file states your preference. The logs show what happened.

OpenAI notes that changes to its search crawler settings can take about 24 hours to work through its systems. That is another reason to plan the change, verify it and monitor the result rather than editing the file blind.

What to check before changing robots.txt

Start with the business decision, then translate it into technical rules. Do not start with a plugin checkbox.

1. Decide whether you want the site eligible for ChatGPT search answers.
2. Make a separate decision about allowing content to be used for model training.
3. Check the live robots.txt, including rules added by WordPress plugins, security tools or a CDN.
4. Review server or CDN logs for OpenAI-labelled requests and verify them against published IP ranges.
5. Test the final file and monitor the site after the change.

Do not hand-edit a live file unless you understand how your WordPress setup generates it. A plugin, hosting layer or CDN may overwrite the change or apply another rule elsewhere.

Make the choice before the software makes it for you

Blocking GPTBot is a training decision. Blocking OAI-SearchBot is a search visibility decision. A rule aimed at ChatGPT-User may not provide the hard block its name suggests.

• Keep the three controls separate.
• Match each rule to a stated business goal.
• Use logs to verify access rather than trusting labels alone.
• Use server or CDN controls when access genuinely must be stopped.

If you want to know what your current setup is allowing, blocking or accidentally hiding, a technical SEO and AI visibility review gives you the answer before anyone touches the file.

The question is not whether every AI bot should be blocked or welcomed. It is whether your website is making the choice you intended.

Source

OpenAI, Overview of OpenAI Crawlers, retrieved 17 August 2026.