What do people actually use ChatGPT for? OpenAI provides some numbers. (arstechnica.com)

OpenAI’s Economic Research team, with Harvard economist David Denning, published a 65‑page NBER working paper that for the first time uses OpenAI’s internal logs (not self‑reported surveys) to map how people actually use ChatGPT across time and tasks. The study provides a rare, large‑scale snapshot of real user behavior and distills seven key findings about adoption, usage patterns and task mix—giving researchers, product teams and policymakers more reliable evidence about where LLMs are being used and how quickly usage is growing. The headline technical takeaways: weekly active users on consumer ChatGPT plans rose from ~100 million in early 2024 to over 400 million earlier this year and now exceed 700 million (OpenAI calls this nearly 10% of the world’s adults), though the company flags measurement noise from device duplication and multiple accounts. Only a small share of users currently pay. Because the paper is based on internal interaction logs, it improves accuracy for temporal and task‑level analyses and has immediate implications for measuring AI’s economic impact, designing product monetization and shaping regulation—while reminding readers to account for known data caveats when interpreting scale and engagement.