A new study from Pew Research, released on Thursday, reveals that over one-third of webpages published since the launch of ChatGPT show signs of being written or substantially edited by AI. This report aligns with other studies detailing the increasing presence of AI in web content creation.
Pew's analysis focused on the content being browsed online rather than web traffic sources. The study utilised the Common Crawl web archive to gather nearly half a million English-language webpages from the past five years, including those published before ChatGPT's November 2022 release. Open Pangram's technology was then employed to identify pages likely written or heavily edited by AI.
When specifically examining webpages published after ChatGPT's release, over one-third (35%) displayed signs of AI authorship. The study also found that .com domains exhibited signs of AI authorship at roughly ten times the rate of .edu or .gov domains, which were both around 1% AI-authored. .org domains had a 4.6% rate of AI authorship.
Pew acknowledged that AI detection tools like Pangram can sometimes misclassify pages, but stated that at scale, the data is likely directionally correct. Additionally, the study noted an increase in other indicators of AI authorship over the years, such as the use of em dashes, Oxford commas, and specific phrasing.