Pew research finds AI influence on over a third of web pages since ChatGPT launched

A Pew Research Center study, released on August 20, found that more than a third of web pages published since the launch of ChatGPT in November 2022 show signs of AI-generated content. The research, which analyzed nearly 500,000 English web pages from the Common Crawl archive, used Open Pangram technology to detect AI involvement. After excluding pages published before ChatGPT’s launch, the study found that 35% of new web pages contained AI-generated content. .com websites were found to have AI traces at a rate about 10 times higher than .edu and .gov sites, with the latter having around 1% AI content. .org sites showed a 4.6% AI presence. The study also noted an increase in writing styles typically associated with AI, such as the use of em dashes, Oxford commas, and phrases like 'not X but Y'. The research acknowledged potential inaccuracies in AI detection tools, but emphasized that the findings reflect broader trends. Meanwhile, Cloudflare reported that bot-generated traffic now exceeds human traffic, a trend that arrived earlier than previously predicted.