Key facts
- Approximately 35% of English-language webpages published since November 2022 show significant signs of AI authorship.
- One in 10 English-language webpages overall exhibit signs of AI authorship.
- .com domains show AI authorship signs at roughly 10 times the rate of .edu or .gov domains.
- AI detection models analyze statistical patterns in text, not specific words.
- Certain linguistic patterns and word choices have become more common in AI-generated text since 2023.
A significant portion of the internet is now being written by artificial intelligence, according to a Pew Research Center study. The analysis found that one in 10 English-language webpages exhibits substantial signs of AI authorship. This figure rises dramatically to 35% when considering only content published after the launch of OpenAI's ChatGPT in November 2022.
The study, which examined approximately 490,000 webpages from the Common Crawl archive using the Open Pangram AI detection model, revealed disparities in AI content prevalence across different domain types. Pages on .com domains showed signs of AI authorship at a rate about ten times higher than .edu or .gov domains, which each registered around 1%. .org sites had a rate of 4.6%.
Researchers identified specific linguistic patterns that have become more common in AI-generated text since 2023. These include an increased use of em dashes and Oxford commas, a more frequent appearance of words like 'delve' and 'interplay,' and a near tripling of 'negative parallelism' constructions. However, Pew cautioned that AI detection models are not infallible and can misclassify pages, and that 'significant signs of AI authorship' does not necessarily mean the content was entirely machine-generated, but could be AI-assisted.
The prevalence of AI-generated content on .com domains, which include news sites and high-volume content farms, is attributed to the rapid publishing cycles that outpace human editorial oversight. In contrast, academic and government sites undergo more rigorous review processes. The trend line for .com AI authorship is steep, climbing from approximately 1% in January 2021 to a projected 9.35% by January 2026.