How People Use ChatGPT and Why You Should Care (kuber.studio)

OpenAI published a large-scale study (1.5M conversations) detailing how people actually use ChatGPT, revealing both the platform’s massive reach and surprising behavioral shifts. By July 2025 ChatGPT hit roughly 700 million weekly active users sending 2.5 billion messages per day (~29,000 messages/second), reaching about 10% of the world’s adult population. Usage is now predominantly personal: only 27% of messages are work-related (down from 47% in mid‑2024) as personal “home production” use exploded. Nearly 80% of interactions fall into three buckets—Practical Guidance (29%), Writing (24%, mainly editing existing text and declining), and Seeking Information (24%, rapidly growing). Coding comprises just ~4.2% of messages. Demographics shifted fast: users with typically feminine names are now a majority, ~50% of messages come from people under 26, and adoption is accelerating in low‑to‑middle‑income countries. The study used automated LLM classifiers and secure clean rooms so no humans read user messages, setting a privacy-preserving research precedent (though classifier accuracy remains a caveat). Key implications for builders and researchers: prioritize personal, decision‑support experiences over pure task automation; treat writing assistance largely as editing tools; invest in better information‑seeking/search interfaces; and design for global, younger, and increasingly gender-balanced audiences. The results signal AI’s evolution from content generator to ubiquitous advisor — reshaping both consumer behavior and the product priorities in ML.