OpenAI reveals biggest-ever study of how people are using ChatGPT – here are 3 things we've learned (www.techradar.com)

OpenAI released the largest-ever study of ChatGPT usage, conducted by its Economic Research team with Harvard economist David Deming for the NBER, detailing who uses the model, what they use it for, and how use is changing. Key findings: 70% of users interact with ChatGPT outside work; nearly half of queries are “Asking” (advice/information), 40% are “Doing” (drafting, planning, programming), and 11% are “Expressing” (reflection/play); three-quarters of conversations target practical guidance, information, or writing. Demographic and geographic shifts are notable: the share of users with feminine names rose from 37% (Jan 2024) to 52% (July 2025), and adoption in the lowest-income countries is occurring four times faster than in the wealthiest. For the AI/ML community these results matter for product design, evaluation, and deployment: the dominance of practical, off‑work use and writing tasks suggests prioritizing reliability, context retention, and editability over novelty; rapid uptake in lower‑income regions raises issues of localization, bandwidth and compute efficiency, and equitable model access; and narrowing demographic gaps affect the representativeness of feedback signals used to fine‑tune models. The breakdown of activity types (Asking/Doing/Expressing) can guide benchmark construction, user-intent detection, safety tuning, and feature prioritization for future model iterations.