The internet is changing, and the evidence is in the text. Recent research from the Pew Research Center analyzed nearly half a million English-language webpages to determine how much online content is generated or edited by artificial intelligence. By using the Open Pangram detection tool on samples from the Common Crawl archive, researchers identified a clear shift in authorship patterns across the web.
Data from early 2026 shows that approximately one-in-ten pages on .com domains now exhibit signs of AI authorship. This rate is significantly higher than on .org domains, which sit at 4.6%, and stands in contrast to .edu and .gov domains, where AI-detected content remains at roughly 1%. The study tracked these trends starting from 2021, showing a steady rise in automated text as generative tools became more accessible to the public.
Beyond simple authorship rates, the study identified specific linguistic quirks that often appear in AI-generated text. These signals include an increased frequency of em dashes, Oxford commas, and specific vocabulary choices like "interplay" or "valuable." While no single punctuation mark confirms the use of AI, the combination of these patterns suggests a measurable shift in how digital information is produced.
This research highlights how AI-driven content is not distributed equally across different types of websites. As the public continues to integrate chatbots into daily routines, the distinction between human and machine expression becomes less obvious. Understanding these trends helps researchers and casual readers alike track how the digital environment is being reshaped by large-scale language models.