67% of frequently cited websites that block GPTBot, OpenAI's training crawler, still appear as sources in ChatGPT answers, close to the 73% of all frequently cited websites. Blocking GPTBot trims a site's ChatGPT citations; it does not remove the site from ChatGPT.
Analysis of 813 frequently cited websites and 222,157 citations · answers from ChatGPT, Gemini, Perplexity and Google's AI features, Google search results and the public HTTP Archive crawl, September 2026
Key Findings
- 1.67% of frequently cited websites that block OpenAI's GPTBot still appear as sources in ChatGPT answers, against 73% of all frequently cited websites.
- 2.22.6 ChatGPT citations on average go to each website that blocks GPTBot, 29% fewer than the 31.8 for websites that allow it.
- 3.70% of frequently cited websites that block OpenAI's search crawlers, OAI-SearchBot or ChatGPT-User, still appear as sources in ChatGPT answers.
- 4.31% of the most-cited websites block at least one AI crawler, and those websites hold 37% of all citations.
- 5.Websites that block an AI crawler average 330 citations each, a third more than the 248 of websites that block none.
Summary
A publisher adds two lines to its robots.txt file, the text file that tells crawlers which pages they may read: "User-agent: GPTBot" and "Disallow: /". The goal is to keep its articles out of ChatGPT. A month later, ChatGPT is still citing those articles in its answers.
That happens because GPTBot is OpenAI's training crawler, the bot that collects pages to train future models. ChatGPT's search uses different crawlers and can find pages through search indexes, so a GPTBot block works on training and barely touches answers. Site owners who block GPTBot to stay out of ChatGPT, and owners worried that blocking will cost them visibility, both need to know what the block actually does.
The answer is that it mostly does not keep a site out. 67% of frequently cited websites that block GPTBot are still cited by ChatGPT, against 73% of all frequently cited websites. Blocking cuts the number of ChatGPT citations a site receives by 29%, and even blocking OpenAI's search crawlers leaves 70% of sites in ChatGPT's answers.
What we measured
On September 23, 2026 we recorded the sources cited in answers from ChatGPT, Gemini, Perplexity, Google AI Mode and Google AI Overviews, together with Google's top 10 search results for the same questions. We kept every website that appeared at least 50 times, 813 websites holding 222,157 citations, and matched each one to its robots.txt file in the public HTTP Archive crawl of September 2026. We then compared how often ChatGPT cites websites that block AI crawlers with websites that do not.
A citation is a source link shown with an answer or a search result. A website blocks a crawler when its robots.txt file includes a rule disallowing it. GPTBot collects training data for OpenAI's models. OAI-SearchBot and ChatGPT-User are OpenAI's search crawlers: the first builds the index ChatGPT searches, the second fetches a page when a user asks ChatGPT to look something up. A frequently cited website is one with 50 or more citations.
Two thirds of GPTBot blockers are still cited by ChatGPT
The gap to websites that allow GPTBot is small. 75% of frequently cited websites that allow GPTBot are cited by ChatGPT, and 73% of all frequently cited websites are. A block moves a site from roughly three in four to two in three: a real difference, but most blocked sites stay in ChatGPT's answers.
Source: Suff Digital analysis of 813 frequently cited websites and 222,157 citations · answers from ChatGPT, Gemini, Perplexity and Google's AI features, Google search results and the public HTTP Archive crawl, September 2026
| Group of frequently cited websites | Websites | Cited by ChatGPT |
|---|---|---|
| All frequently cited websites | 813 | 73% |
| Block GPTBot | 176 | 67% |
| Allow GPTBot | 637 | 75% |
| Block OAI-SearchBot or ChatGPT-User | 127 | 70% |
Blocking OpenAI's search crawlers does not change the picture much either. 70% of frequently cited websites that block OAI-SearchBot or ChatGPT-User still appear in ChatGPT citations, 89 websites in all. ChatGPT can reach a page through other search indexes and through links elsewhere on the web, so a robots.txt rule limits what OpenAI's own crawlers fetch, not whether a page can be cited.
Blocking GPTBot lowers ChatGPT citations by 29%
The block does have an effect on volume. Websites that block GPTBot receive 22.6 ChatGPT citations each on average, against 31.8 for websites that allow it, 29% fewer. A blocked site keeps its place in ChatGPT's answers but shows up in fewer of them.
For a publisher, that is the trade-off in plain numbers: a GPTBot block keeps content out of OpenAI's model training at the cost of roughly three in ten ChatGPT citations. For a business that wants to be recommended, that cost buys nothing it needs.
The websites that block AI are the ones AI cites most
31% of the most-cited websites block at least one AI crawler, and they hold 37% of all citations. Across every AI engine and Google's results, websites that block an AI crawler average 330 citations each against 248 for websites that block none, and their median is 133 against 97.
Source: Suff Digital analysis of 813 frequently cited websites and 222,157 citations · answers from ChatGPT, Gemini, Perplexity and Google's AI features, Google search results and the public HTTP Archive crawl, September 2026
| Measure | Block at least one AI crawler | Block no AI crawler |
|---|---|---|
| Websites | 250 | 563 |
| Average citations per website | 330 | 248 |
| Median citations per website | 133 | 97 |
| Share of all citations | 37% | 63% |
This reflects size. News outlets and major publishers are the websites most likely to block AI crawlers, and they are also the sources AI engines and Google most want to cite. Among the 250 blockers, 156 block Anthropic's ClaudeBot, 156 block Common Crawl's CCBot, 120 opt out of Google's AI training with Google-Extended and 119 block PerplexityBot.
What this means for publishers and brands
If you block GPTBot to stay out of ChatGPT answers, the block is doing a different job. It keeps your pages out of OpenAI's model training, and it trims your ChatGPT citations by about 29%, but two in three blocked websites are still cited. If staying out of model training is your goal, a GPTBot block is the right tool, and it works without removing you from answers.
Most businesses want the citations. For them, the right setting is to allow OAI-SearchBot and ChatGPT-User, make a deliberate choice about training crawlers such as GPTBot and Google-Extended, and check that robots.txt reflects that choice. Suff Digital is an ai seo company that reviews crawler rules for exactly this trade-off and tracks which engines cite your site.
Embed this research
Paste this on your site to embed the charts. It links back to the source automatically.
<iframe id="sd-gptbot-blocking-chatgpt-citations" src="https://www.suffdigital.com/embed/data-studies/gptbot-blocking-chatgpt-citations" width="100%" height="600" style="width:100%;border:1px solid #E5E7EB;border-radius:12px" loading="lazy" title="Blocking Doesn't Keep You Out: 67% of Websites That Block GPTBot Still Get Cited by ChatGPT - Suff Digital"></iframe>
<script>window.addEventListener("message",function(e){if(e&&e.data&&e.data.sdEmbed==="gptbot-blocking-chatgpt-citations"&&e.data.height){var f=document.getElementById("sd-gptbot-blocking-chatgpt-citations");if(f){f.style.height=e.data.height+"px";}}});</script>
<p style="font:14px/1.5 system-ui,sans-serif">Source: <a href="https://www.suffdigital.com/resources/data-studies/gptbot-blocking-chatgpt-citations">Blocking Doesn't Keep You Out: 67% of Websites That Block GPTBot Still Get Cited by ChatGPT - Suff Digital</a></p>
Cite this study
Suff Digital. (2026). Blocking Doesn't Keep You Out: 67% of Websites That Block GPTBot Still Get Cited by ChatGPT. https://www.suffdigital.com/resources/data-studies/gptbot-blocking-chatgpt-citations
Blocking Doesn't Keep You Out: 67% of Websites That Block GPTBot Still Get Cited by ChatGPT - Suff Digital - https://www.suffdigital.com/resources/data-studies/gptbot-blocking-chatgpt-citations
Frequently asked questions
Related studies
- AI SEO1.5 Million Websites Block at Least One AI Crawler From Reading Their Pages
- AI SEOThe 1,000 Most Visited Websites Block Perplexity's AI Crawler 12 Times as Often as Small Sites
- AI SEOBrands That Block GPTBot Still Get Named by ChatGPT: 31.5 Mentions vs 20.7 for Brands That Allow It
- AI SEO47% of Top News Sites Block ChatGPT's Training Crawler vs 17.6% of All News Sites