One in three web pages published since ChatGPT launched now shows signs of AI authorship
A Pew Research study of nearly half a million web pages found that 35% of those published after November 2022 carry signs of AI writing. Meanwhile, bots are also doing most of the browsing.
Key points
- Pew Research found that 35% of English-language web pages published after ChatGPT's November 2022 launch show signs of AI authorship, according to a report released on Thursday.
- In a random snapshot of 10,000 pages collected in July 2026, around 10% showed significant AI-writing signals, rising to 35% once pre-ChatGPT pages were removed.
- Commercial .com domains showed AI-authorship signals at roughly ten times the rate of .edu and .gov domains, which both sat near 1%.
- Cloudflare, the company that routes much of the world's internet traffic, separately reported that automated bot traffic has overtaken human browsing traffic.
- AI detection tools can misidentify human writing as AI-generated, so individual results carry uncertainty, though the overall trend in the Pew data is clear.
Most of us assume that when we load a web page, a person wrote it. That assumption is getting harder to hold.
Pew Research released a study on Thursday showing that more than one third of English-language web pages published after the launch of ChatGPT, the AI chatbot that went public in November 2022, show signs of being written or heavily edited by AI. The study, first reported by TechCrunch AI, analysed nearly half a million pages pulled from the Common Crawl web archive, a freely available collection of billions of web pages.
Pew used detection technology from a company called Open Pangram to score each page for AI-writing signals. A random sample of 10,000 pages collected in July 2026 showed about 10% with significant AI-authorship signs. Once researchers filtered out older pages, ones that pre-date AI writing tools and therefore could not have been AI-written, that figure climbed to 35%.
Who is writing, and who is reading?
Bots, increasingly, on both counts. Internet infrastructure company Cloudflare, which handles traffic for a huge share of the world's websites, recently reported that automated bot traffic now outnumbers genuine human browsing. Pew's study does not cover who is visiting pages; it looks at what those pages contain. Put the two findings together and a strange picture emerges: automated programs browsing pages that automated programs wrote.
Not all corners of the web look alike. Commercial .com domains showed AI-writing signals at around ten times the rate of .edu (educational) and .gov (government) sites, both of which came in near 1%. Non-profit .org domains sat at 4.6%. That gap makes sense: AI writing tools lower the cost of producing content at scale, and commercial sites have the strongest financial motive to do exactly that.
How reliable are these numbers?
Partially, but directionally solid. AI-detection tools, including Pangram's, can make mistakes. A human writer who favours short, clean sentences might score as AI-generated. Conversely, a polished AI output could slip through. Pew acknowledged this openly. At nearly half a million pages, though, individual errors average out, and the trend is almost certainly real.
Pew also noted that classic signals of AI writing have grown more common across the web over the study period: frequent use of em dashes (the long dash character AI models favour), Oxford commas, and stock phrases like "it's not X, it's Y."
For ordinary readers, none of this means every page you visit is AI-written. But it does mean the odds have shifted. A product review, a how-to article, a news summary you find via search: any of them may have been assembled by software rather than a person who did the work.
Watch for: pages that sound fluent but feel generic, that cover a topic without any specific detail or personal experience, or that use the same few sentence structures throughout. Those are the fingerprints Pew's study is picking up, one page at a time.
