How Much of the Internet Is Written with AI? (www.pewresearch.org)

A recent analysis of nearly half a million English-language webpages reveals that around 10% exhibit significant signs of AI authorship as of July 2026, marking a notable increase from earlier years post-ChatGPT's release. The study utilized the Common Crawl web archive and an AI detection tool called Open Pangram, which analyzes patterns in language to identify AI-generated content. The findings suggest that the prevalence of AI-written material is particularly pronounced in newer websites, where over one-third of pages published after ChatGPT's introduction show AI characteristics. This shift reflects a broader trend toward the integration of AI-generated text in digital content creation. The significance of this trend for the AI/ML community lies in the implications for content authenticity and influence on the internet. The analysis highlights that while AI authorship is on the rise, its distribution is uneven across different types of domains. For example, AI-authored text appears at significantly higher rates in commercial domains (.com) compared to educational (.edu) and governmental (.gov) sites. Additionally, certain linguistic patterns commonly found in AI-generated text, such as the frequent use of em dashes and Oxford commas, are becoming more prevalent. These trends raise critical questions about the nature of authorship, the evolution of online communication, and the ongoing challenges of identifying AI-generated content accurately.