Data scientists at the Pew Research Center ran AI-detection tools across 500,000 English-language web pages. Their conclusion: AI has already seeped into the capillaries of the internet.

The share grew steadily. After ChatGPT launched in late 2022, the proportion of AI-generated content climbed. By July 2026, more than a third of pages published after ChatGPT's release showed clear traces of AI authorship.

AI content is spread unevenly across domains. When ChatGPT first appeared, AI language patterns showed up at similar rates on .com, .org, .edu, and .gov sites. By 2026, one in ten pages on .com domains had AI traces—twice the rate of .org (4.6%) and ten times the rate of .edu and .gov (1%). Commercial sites are the hardest hit, while academic and government sites remain relatively restrained.

Even more interesting is AI's impact on language habits. Compared with 2023, em dashes now appear twice as often on today's web pages. Oxford commas are up 63%. AI-favorite words like "delve," "interplay," and "testament" have more than doubled in frequency. And contrastive constructions like "it's not just X, it's Y" have nearly tripled. These fingerprints reveal the preferences of generative models—and are quietly rewriting how English is written.

AI-detection tools themselves aren't perfect, but the trend is clear. When AI-generated content becomes so abundant that it shifts statistical patterns, the line between "written by a human" and "written by a machine" on the internet gets blurrier by the day. Next time you read an article full of em dashes and the word "delve," take a guess: is there a language model behind it?