Pew Study Finds AI Writes Over a Third of Post-ChatGPT Web Pages
A new Pew Research study finds that more than one-third of web pages published since ChatGPT's November 2022 launch show signs of being written or heavily edited by AI. Pew collects nearly half a million English-language web pages from the Common Crawl archive and uses Pangram's detection technology to analyze them. In a random July 2026 sample of 10,000 pages, about 10% show significant AI authorship, but that figure includes older pages published before AI writing tools existed.
When Pew filters out older pages and focuses only on those published after ChatGPT's release, the share with signs of AI authorship jumps to 35%. The study also finds striking differences across domain types: .com pages show AI authorship at roughly ten times the rate of .edu and .gov domains, which both sit near 1%, while .org pages come in at 4.6%. Pew additionally notes an increase in common AI stylistic markers, such as em dashes, Oxford commas, and phrasing like "it's not X, it's Y."
The report adds to growing evidence of a web increasingly shaped by machines, arriving shortly after Cloudflare reports that bot traffic now exceeds human traffic. Pew's data shifts the focus from who is browsing the web to what is being browsed, suggesting bots are often reading pages written by other bots. The analysis has limitations, since AI-detection tools like Pangram can misclassify human-written pages, but at scale the data is likely directionally correct.