AI Is Flooding New Web Pages, but the Detectors Still Have Limits

AI Is Flooding New Web Pages, but the Detectors Still Have Limits
The web’s newest pages are increasingly being made by machines, not merely read by them. A Pew Research analysis suggests the shift accelerated sharply after ChatGPT’s arrival, while leaving a stubborn question over how precisely AI authorship can be measured.

The turning point in the study is November 22, 2022, when ChatGPT was released. Looking across nearly half a million English-language pages preserved by Common Crawl over roughly five years, Pew and AI-detection firm Open Pangram found that older pages showed relatively few signs of machine involvement. But among pages published after ChatGPT launched, more than one-third were likely written or substantially edited by AI.

In a July 2026 random sample of 10,000 pages, about 10% showed significant signs of AI authorship. That headline number obscures the generational divide: the sample included material published before widely available AI writing tools. Once Pew excluded those older pages, the apparent AI share rose to 35%.

The pattern was not evenly spread. Commercial .com addresses showed signs of AI authorship at roughly 10 times the rate of .edu and .gov sites, which each registered around 1%; .org pages stood at 4.6%. Pew also recorded a rise in stylistic signals often associated with generative AI, including em dashes, Oxford commas and the construction, “it’s not X, it’s Y.”

The sources agree on the direction of travel: the open web is increasingly saturated with AI-generated prose. Yet the more cautious account matters too. Detection tools can wrongly label human writing as AI-made, meaning the figures are an estimate rather than a final audit. Even so, the study’s central picture is hard to miss: newer web content is being reshaped at industrial speed, in a system where bots may increasingly consume pages produced by other bots.

Continue reading https://foxvector.com/stories/01a032e3-a3ae-1667-7016-316f7a4914fe