A large-scale study has revealed a rapid increase in the presence of AI-generated content online since 2022, reports infohub.kz.
The Pew Research Center analyzed about 490,000 English-language web pages collected from January 2021 to July 2026. Of the random sample of pages published after the launch of ChatGPT in November 2022, 35 percent showed clear signs of AI use. Search services and text models such as ChatGPT, Google Gemini, and Anthropic Claude are having an increasingly noticeable impact on published content.
The highest concentration of generated text was found on commercial sites. Roughly one in ten pages in the .com domain zone shows signs of AI, while for .org domains the figure was 4.6 percent, and for .edu and .gov it did not exceed one percent. Before the advent of popular neural networks, the presence of machine text on all types of sites was within one percent.
Analysts also noticed a shift in online language habits. Since the beginning of 2023, the use of the em dash has doubled, and the Oxford comma has increased by 63 percent. Additionally, the use of words characteristic of neural networks, such as "delve" and "testament," has more than doubled. However, the study authors emphasize that these signs may reflect both algorithmic work and ordinary human editing.