Can you block GPTBot and still show up in ChatGPT?

What do GPTBot, OAI-SearchBot, and ChatGPT-User each do?

GPTBot collects content for training OpenAI's models, OAI-SearchBot decides which sites can show up in ChatGPT search answers, and ChatGPT-User visits a page when a person asks ChatGPT to do something with it. Each one answers to its own line in robots.txt.

OpenAI's page on its crawlers treats them as separate user agents, so a rule written for one does not touch the others. ChatGPT-User is the odd one out. OpenAI says robots.txt rules may not apply to it, since a person started the visit, and that it is not used to decide whether content may appear in search.

The same page has a footnote that is easy to skip. Sites opted out of OAI-SearchBot will not be shown in ChatGPT search answers, OpenAI says, though they can still appear as navigational links.

Which should you block or allow?

If you want out of training and still want to be found, disallow GPTBot and allow OAI-SearchBot. OpenAI's documentation names that exact combination as an option.

In robots.txt that takes two short groups. The first reads User-agent: GPTBot followed by Disallow: /, and the second reads User-agent: OAI-SearchBot followed by Allow: /. Nothing needs writing for ChatGPT-User, since OpenAI says the file may not govern it anyway.

Blocking OAI-SearchBot is a separate decision, and it is the one that keeps your pages out of ChatGPT search answers. We would only make it if being absent from those answers is the point. How ChatGPT chooses among the sites it can read is a longer question, which our guide to AI search covers.

How long does a robots.txt change take?

About a day, by OpenAI's own estimate. Its documentation says it can take roughly 24 hours from a robots.txt update for its systems to adjust.

That covers the crawler picking up the new rule. It says nothing about when answers will reflect it, and answers shift on their own from week to week, which our post on how often AI answers change goes through.

In our AI search work we take a baseline on day one and report your share of AI answers against it every month, so a robots.txt change can be read as a before and after.

Do blocked sites still get cited?

Some do. In BuzzStream's study of the top 50 news sites, most of those blocking OpenAI's crawlers still appeared in AI citations.

The study, reported by PPC Land in March 2026, looked at 4 million citations from 3,600 prompts across 10 industries in ChatGPT, Gemini, Google AI Overviews and AI Mode. Its blocking analysis covered only the top 50 news sites. Of those blocking OAI-SearchBot, 82.4 percent still appeared in citations, and of those blocking GPTBot, 88.2 percent did.

Go back to that footnote on OpenAI's page. A site opted out of OAI-SearchBot can still turn up as a navigational link, and this study counted citations across four AI tools, only one of them OpenAI's. BuzzStream draws no causal conclusions, and we would not either.

Two limits apply before anyone reads this as a rule for their own site. News publishers get cited for reasons a B2B software company does not, and OpenAI still says that blocking OAI-SearchBot keeps a site out of ChatGPT search answers.

How common is blocking?

Common among news publishers and rare across the wider web. HasData found that 50.5 percent of news publishers block GPTBot, against 7.9 percent of top-web sites.

HasData's index of 10,894 domains mixed the most visited sites on the web with news publishers. A second count points the same way. A Cloudflare Radar snapshot of 4,242 robots.txt files on 28 September 2026, as reported by Technology Checker, found GPTBot in 520 disallow rules against 326 allow rules. OAI-SearchBot came out the other way round, with 268 disallow rules and 295 allow rules.

HasData also tested whether the bans hold. Of 592 sites in its test that disallow GPTBot, 234, or 39.5 percent, still served the crawler a live page. A line in robots.txt is a request, and in those cases the sites' own servers did not act on it.