Introduction
A new study from the Pew Research Center has found that more than one-third of web pages published since the launch of ChatGPT show signs of having been written or substantially edited by artificial intelligence. The report, titled "How Much of the Internet Is Written With AI?," analyzed almost half a million English-language web pages collected from the Common Crawl web archive, including material published before generative AI tools became widely available.
Key Findings
According to the study, in a random sample of 10,000 web pages collected in July 2026, around 10% showed significant signs of AI authorship. The sample included pages published over the past five years, as well as older content dating back to the early days of the web. TechRadar reported that 9.6% of all pages in the sample appeared to have been written by AI, a figure that reflects the inclusion of pages published more than 30 years ago.
The analysis found that the share of web pages showing signs of AI authorship has risen steadily since ChatGPT's arrival in November 2022. The study's authors note that the results should be interpreted as evidence of AI authorship rather than definitive proof, given the limitations of detection methods.
Domain Patterns
Both TechRadar and Times Now reported that AI-written content is more common on certain types of domains. In the 2026 samples, web pages with a .com domain showed signs of AI authorship at a rate of about 9.4%, compared with 4.6% for .org domains. Educational (.edu) and government (.gov) domains showed much lower rates, at around 1% each. TechRadar noted that the majority of suspected AI-generated web pages published in 2026 are .com domains, which are typically commercial, and that it is more likely that .edu or .gov pages are written by humans.
Linguistic Markers
Pew's analysis, which used a machine learning model called Open Pangram to examine linguistic patterns, identified several distinctive features of AI-generated text. The em dash appears twice as frequently in pages published since 2023, and the use of Oxford commas has increased by 63%. Words like 'delve', 'interplay', and 'testament' have more than doubled in usage, while negative parallelism — a rhetorical structure — has almost trebled in use.
Context and Broader Trends
The findings come as AI tools have become increasingly embedded in online content creation. Times Now reported that around half of U.S. adults say they use AI chatbots, including 24% who use them daily. The report also cites broader trends in internet traffic: Cloudflare CEO Matthew Prince said in June that bot traffic overtook human traffic online, driven by AI agents. While the study focuses on written content, the rise of AI-generated pages has implications for how users and search engines distinguish between human and machine-produced information.
Perspectives
Pew Research Center
The study's authors emphasize that the findings are evidence of AI authorship, not definitive proof, and highlight that the share of AI-written pages has risen steadily since ChatGPT's launch. They also note that AI-written content is more prevalent on commercial domains, which may reflect the use of AI for marketing and business purposes.
Cloudflare CEO Matthew Prince
In a separate observation, Matthew Prince, CEO of Cloudflare, said that bot traffic overtook human traffic online in June, driven by AI agents. This comment, reported by Times Now, underscores the broader impact of AI on internet activity beyond just written content.