A recent study by Pew Research, published Thursday, reveals that over a third of all webpages created since the launch of ChatGPT exhibit characteristics of AI authorship. This finding reinforces prior research indicating a significant portion of newer web content is either generated or extensively revised by artificial intelligence, according to the firm.
This report emerges soon after Cloudflare, an internet infrastructure giant, announced that automated bot traffic now exceeds human internet activity – a benchmark achieved ahead of their initial projections.
However, Pew's analysis distinguishes itself by focusing not on the identity of web browsers, but rather on the nature of the content being consumed. Intriguingly, it suggests a scenario where a considerable amount of web content is being accessed by bots that were themselves generated by other AI systems.
For this comprehensive report, Pew utilized the Common Crawl web archive to gather approximately half a million English-language webpages spanning the last five years, commencing a few years prior to ChatGPT's debut in November 2022. The firm then employed Open Pangram's specialized technology to identify pages likely authored or substantially modified by AI.
A preliminary random sample of 10,000 webpages, collected in July 2026, indicated that roughly 10% displayed “significant signs of AI authorship,” according to Pew.
Pew acknowledged, however, that such a broad random sample would inherently include older webpages published before the advent of AI writing tools, thus making AI authorship impossible for those specific entries.
To gain a more accurate understanding of current AI-generated web content, Pew subsequently filtered out these older pages, concentrating solely on those published after ChatGPT's introduction.
Within this refined dataset, with pre-AI pages excluded, the proportion of pages exhibiting AI authorship surged to over one-third, specifically 35%.
Further granular analysis by domain revealed a significant disparity: .com URLs displayed signs of AI authorship at approximately ten times the rate of .edu or .gov domains, both of which registered around 1% AI authorship. Furthermore, .org domains showed a comparatively lower rate of 4.6%.
Naturally, this analytical approach is not without its limitations. Pangram, similar to other AI detection tools, may occasionally misclassify human-written content as AI-generated. Nevertheless, when applied at scale, the data is considered to be directionally accurate.
Additionally, Pew identified an increase over time in other stylistic indicators often associated with AI authorship, including the frequent use of em dashes, Oxford commas, and specific rhetorical structures such as “it's not X, it's Y.”
The Editorial Staff at AIChief is a team of professional content writers with extensive experience in AI and marketing. Founded in 2025, AIChief has quickly grown into the largest free AI resource hub in the industry.