Should you block GPTBot? A plain answer

Blocking GPTBot is a choice about AI training, not about ChatGPT search. Per BotView's guide, checked October 21, 2026, blocking GPTBot opts you out of model training only and does not remove you from ChatGPT search, which a different crawler, OAI-SearchBot, controls. Unveilr's comparison reaches the same conclusion. So block GPTBot if you object to your pages being used for training, and leave OAI-SearchBot alone if you want to stay eligible for ChatGPT search answers.

Quick version

• GPTBot gathers content for training OpenAI's models. OAI-SearchBot indexes pages for ChatGPT search (ICODA's guide).
• You can block one and allow the other. They are separate rules in robots.txt (BotView).
• Blocking GPTBot has no effect on Google (ICODA's guide).
• The costly mistake is blocking OAI-SearchBot by accident, usually with an old file or a wildcard rule.
• If you only want the lines to paste, our short guide on blocking ChatGPT in robots.txt has them. The rest of this article explains the decision first.

What each OpenAI bot does

OpenAI publishes one page that describes its robots, with a separate user-agent string and a separate published IP list for each. The table below summarizes what the sources we checked on October 21, 2026 say about each one.

Two of those sources sell related services. ICODA offers an AI visibility tool that checks robots.txt and crawler access. Eight25 Media sells enterprise CMS migration services. Both match what the other sources say here, but OpenAI changes crawler details over time. BotView notes that OpenAI increments the GPTBot version number without notice, so check OpenAI's own published crawler documentation before you change a live file.

Should you block GPTBot?

The useful question is not whether blocking helps your search presence. Unveilr's comparison says only OAI-SearchBot decides whether your site can appear in ChatGPT search answers, so blocking GPTBot alone costs you nothing in visibility. The question is whether you are comfortable with your content being used to train models.

Our view: this is a values and business decision, and the sources do not settle it for you. This is our judgment, not a sourced fact.

Blocking GPTBot makes sense if:

• Your content is the product, such as paid reports, courses or original research, and you do not want it feeding a model.
• You have a stated policy against AI training use.
• You would rather opt out now and revisit later.

Leaving GPTBot allowed makes sense if:

• You have no objection to training use and want the simplest robots.txt.
• Your content is public marketing material that you already want widely read.

One limit to know about. Unveilr says a Disallow rule under GPTBot opts your future content out of training while leaving search and user fetches untouched. The word to notice is future. We found nothing in the sources saying a block removes what was already collected.

The robots.txt lines to use

To block training but stay in ChatGPT search, Crawloria's recommended setup uses separate rules for the two bots. Crawloria sells bot audits and crawler analysis, so treat it as a vendor view, though BotView and Unveilr describe the same split. Here is the file:

User-agent: GPTBot
Disallow: /

User-agent: OAI-SearchBot
Allow: /

Per BotView, the two lines for GPTBot alone are User-agent: GPTBot and Disallow: /, placed in the robots.txt at the root of your domain. The same guide says path rules let you block part of a site. For example, Disallow: /private/ blocks GPTBot from that directory and allows the rest.

If you do not want to block anything, you do not need any GPTBot line at all.

Mistakes that block search by accident

Most of the damage here comes from blocking the wrong bot, not from the GPTBot decision itself.

1. A wildcard rule. Unveilr says a staging site copied to production with a block-everything rule is one of the most common reasons a brand vanishes from ChatGPT search overnight.
2. An old file. ICODA points out that most guides from 2023 and 2024 only mention GPTBot, because OAI-SearchBot did not exist yet. A file untouched since then may be silently blocking the search crawler.
3. A firewall in front of the site. BotView says your robots.txt can allow GPTBot while Cloudflare, DataDome, Akamai or Imperva quietly serves it a challenge page or an Access Denied page.
4. Blocking OAI-SearchBot on purpose without knowing the cost. Crawloria says doing so makes you invisible in ChatGPT Search, but does not prevent training data collection. Unveilr adds that opted-out sites can still appear as navigational links.

After any change, wait. Crawloria says OpenAI's systems take roughly 24 hours to pick up a robots.txt change.

What other sites are doing

Blocking GPTBot is common, and blocking the search bot is much less so. Eight25 Media reports that among 7,870 top domains studied, 17.6% blocked GPTBot while 6.6% blocked OAI-SearchBot.

Defaults are shifting too. According to Eight25 Media, Cloudflare changed the default for new domains in July 2025 so AI training crawlers were blocked unless the owner chose otherwise. On September 15, 2026, it introduced more granular Search, Training and Agent controls, with Training and Agent traffic blocked by default on pages displaying ads for new domains. If your site sits behind Cloudflare, check what is set there as well as in robots.txt.

What the traffic data does and does not show

Two vendor studies are worth reading with caution. Neither promises anything for your site.

• Crawloria reports that ChatGPT was the number 2 referrer of human traffic to crawloria.com, behind only Google (GA4, June 2026). That is one site's analytics, from a company that sells crawler audits. It does not predict what you would see.
• Presenc AI reports that in a 24-hour window, GPTBot made 222 requests while OAI-SearchBot made 56, a 4:1 ratio. The data came from 300 new programmatic pages on presenc.ai during March 10 to 13, 2026. Presenc AI sells crawler monitoring tools, and this is a small test on one site.

The same study found only 26 pages, 20.5% of crawled pages, were visited by both bots (Presenc AI). That fits the idea that the two bots run as separate systems. It says nothing about whether being crawled leads to being cited.

What about Google?

Blocking GPTBot has no effect on Google. ICODA says OpenAI and Google run completely separate crawlers and separate systems. Google has its own training control. Eight25 Media says Google-Extended controls whether Googlebot-crawled content can train future Gemini models, and that Google states it has no effect on Search inclusion or ranking.

How to confirm the bot you are seeing is real

Do not trust the user-agent string alone. BotView says the only reliable check is the source IP. OpenAI publishes the ranges each crawler uses as JSON, and a request claiming to be GPTBot from outside gptbot.json is forged. ICODA lists the files at openai.com/gptbot.json for GPTBot and openai.com/searchbot.json for OAI-SearchBot.

Match on the token, not the version. BotView says OpenAI increments the version without notice, so an exact match silently stops working. ICODA gives GPTBot's current user agent as GPTBot/1.4, and Crawloria shows OAI-SearchBot/1.3 in its example string, which it notes changes over time.

To see what your own robots.txt currently allows for these bots, our free AI visibility check reads it for you. It reports what is allowed. It does not predict citations or traffic.

The short decision rule

1. Decide whether you object to training use. If yes, add the GPTBot block.
2. Either way, make sure OAI-SearchBot is not blocked unless you want out of ChatGPT search.
3. Check for wildcard rules and firewall settings that override your intent.
4. Wait about 24 hours and recheck.