More than one-third of English-language web pages published after ChatGPT launched show significant signs of AI authorship or editing, according to a new Pew Research Center study. Pew found AI-related signals in 35% of pages published after November 2022, adding to evidence that generative AI is increasingly shaping newly published material on the web.
The report follows recent findings from Cloudflare that automated bot traffic has overtaken human traffic on the web. Pew’s research examined the content being published rather than who or what was accessing it.
Pew Analyzed Nearly Half a Million Web Pages
Pew used the Common Crawl archive to collect nearly 500,000 English-language web pages spanning roughly five years, including material published before and after ChatGPT’s November 2022 release.
Researchers then used technology from Open Pangram to estimate which pages were likely written or substantially edited using AI.
In a random sample of 10,000 pages collected in July 2026, about 10% showed significant signs of AI authorship. Pew noted that this sample also contained older pages published before widely available generative AI writing tools existed.
When researchers removed those older pages and looked only at material published after ChatGPT’s launch, the rate rose to 35%.
Commercial Domains Showed Higher AI Signals
The study found substantial differences between domain types. Pages using .com addresses showed signs of AI authorship at roughly 10 times the rate found on .edu and .gov domains.
Both .edu and .gov pages had AI authorship rates of about 1%. Pages on .org domains showed a higher rate of 4.6%, but still remained well below commercial websites.
Pew cautioned that AI-detection systems are not perfectly accurate. Tools such as Pangram can classify human-written text as AI-generated or fail to identify some AI-written material, so the findings should be treated as estimates rather than exact measurements.
Writing Patterns Have Also Changed
Pew also tracked writing characteristics commonly associated with AI-generated text and found that some became more frequent over time.
Those patterns included increased use of em dashes, Oxford commas, and sentence constructions such as “it’s not X, it’s Y.” None of those features proves that a page was written with AI, but Pew found that their use increased alongside the broader rise in detected AI authorship.
The study adds to other research examining how much newly published online content is now produced or heavily edited using generative AI systems.
Featured image credits: Wikimedia Commons
For more stories like it, click the +Follow button at the top of this page to follow us.