The Impact of AI-Generated Text on the Internet (ai-on-the-internet.github.io)

A new study highlights the rapid rise of AI-generated and AI-assisted text on the internet, estimating that by mid-2025, around 35% of new websites will fall into this category, a dramatic change since the release of ChatGPT in late 2022. The research, leveraging data from the Internet Archive's Wayback Machine, reveals that while AI text prevalence correlates with a decrease in semantic diversity and an increase in positive sentiment, it does not significantly affect factual accuracy or stylistic diversity. This finding contrasts with common public perceptions of AI's negative impact on online discourse. The study employed a robust sampling method to analyze how substantial AI text generation reshapes internet content, surveying 853 US adults about their experiences and concerns regarding AI. Most respondents believed in various negative effects of AI-generated text, including diminished originality and increasing uniformity in writing; however, only two hypotheses were statistically substantiated by the data. This research not only provides insight into the emerging landscape of online content but also sheds light on societal attitudes toward AI's role in shaping communication, suggesting that the implications of AI-generated text are more nuanced than often assumed.