One-Third of Recent Web Content Shows AI Authorship

Pew Research analyzed nearly half a million English-language pages using Common Crawl data, employing Open Pangram technology to identify AI-driven patterns. While a broad sample of web pages showed a lower baseline of AI influence, the percentage surged once researchers isolated content produced after November 2022. This trend aligns with recent findings from Cloudflare, which reported that bot-driven traffic now exceeds human activity, suggesting a web increasingly populated by automated content consumed by automated systems.

The distribution of this content is heavily skewed by domain type. Commercial .com sites are ten times more likely to feature AI-authored text compared to .edu or .gov portals, which maintained an AI-authorship rate of roughly 1%. Non-profit .org domains followed at 4.6%. Beyond the technical detection, researchers noted a rise in specific linguistic markers often associated with large language models, including increased usage of Oxford commas, em dashes, and repetitive structural phrasing. While AI-detection tools are prone to occasional misclassification, the study provides a clear directional map of how generative technology is rapidly colonizing the digital landscape.