More than one-third of web pages published since ChatGPT was released have signs of being written by AI, according to a new study released Thursday by Pew Research. The company says the report supports other studies detailing how many of the web’s new web pages are currently being written or “heavily edited” by AI.
It also comes shortly after internet infrastructure provider Cloudflare reported that bot web traffic has surpassed human web traffic. This milestone was achieved sooner than the company expected.
But Pew’s data focuses on what people are browsing the web, not who or what they are browsing. And apparently a lot of it is bots reading web pages written by other bots.
To compile the report, Pugh said he used the Common Crawl Web Archive to collect nearly 500,000 English-language web pages from the past five years or so, starting several years before ChatGPT’s November 2022 release. Pew then used Open Pangram’s technology to detect how much of it might have been written or heavily edited by AI.
In a random sample of 10,000 web pages collected in July 2026, about 10% showed “significant signs of AI authorship,” Pew said.
However, Pugh pointed out that such a random sample would necessarily include old webpages, i.e. webpages published before AI writing tools existed, and therefore could not have been written by AI.
Image credit: Pew Research
To better understand how much of the web is currently being written by AI, Pew excluded older web pages and focused only on web pages published after the release of ChatGPT.
In this same snapshot where old pages were removed, we found traces of the AI author on more than a third (35%) of the pages.
Digging into the domains themselves, Pew found that URLs containing .com domains were about 10 times more likely to show signs of AI creation than .edu or .gov domains (both were about 1% AI-created). Moreover, the percentage of AI authors in .org domains was only 4.6%.
Of course, this analysis is not perfect. Pangram, like other AI detection tools, can misclassify pages that are not created by AI. But at scale, the data seems at least directionally correct.
Additionally, Pugh found that other expressions attributed to AI have also increased over the years, including the use of em dashes, the Oxford comma, and expressions like “It’s Y, not X,” among others.
If you buy through links in our articles, we may earn a small commission. This does not affect editorial independence.
