Pew: One-Third of Post-ChatGPT Webpages Show Signs of AI Authorship
What happened: Pew Research Center analyzed nearly 490,000 English-language webpages and found that 35% of those published after ChatGPT's November 2022 launch show significant signs of AI authorship.
What happened: Pew Research Center analyzed nearly 490,000 English-language webpages and found that 35% of those published after ChatGPT's November 2022 launch show significant signs of AI authorship. The study used the Open Pangram AI detector and found the highest concentration on .com domains (~10%), with much lower rates on .edu and .gov sites. Linguistic markers such as increased use of em dashes and Oxford commas were prevalent.
Why it matters: The findings highlight the rapid proliferation of AI-generated or AI-assisted content across the web, especially since late 2022. However, Pew cautions that "significant signs of AI authorship" do not necessarily mean fully machine-generated text, as many pages may be the product of human-AI collaboration. The study underscores both the growing impact of large language models and the challenges in accurately measuring their footprint.
Source: Decrypt