A new Pew Research study released Thursday found that over one-third of web pages published since ChatGPT’s late 2022 release showed signs of AI authorship, corroborating other research on AI’s growing footprint online. Using the Common Crawl archive and detection technology from Open Pangram, Pew analyzed nearly half a million English-language pages, finding that .com domains showed AI authorship at roughly ten times the rate of .edu or .gov sites, while other markers like em dashes and Oxford commas also increased over time. The findings arrived shortly after Cloudflarereported that bot web traffic had overtaken human traffic sooner than expected.
In addition to this, OpenAI launched an Apple Messages plug-in for ChatGPT, allowing users to sort, analyze, edit, or draft messages directly through the assistant, with functionality extending to Codex and ChatGPT Work. The company told TechCrunch that the plugin runs locally, does not fully index user messages, and requires explicit requests before ChatGPT reads message content, though it cautioned users against enabling persistent approval for sent messages.
Meanwhile, data from corporate card provider Ramp showed OpenAI gaining ground against Anthropic among more than 70,000 American business users, with OpenAI growing faster quarter-to-date after Anthropic had led since May. Ramp economist Ara Kharazian attributed OpenAI’s momentum partly to strong developer reception for its newer model, while noting Anthropic’s higher-end Fable tier faced adoption friction tied to pricing and data retention requirements. Kharazian noted the overall share of Ramp customers paying for AI tools had climbed to nearly 56 percent by July, suggesting both companies continue expanding even as market share fluctuates between them.
