How people are using ChatGPT (openai.com)

OpenAI and Harvard economist David Deming released the largest study to date of consumer ChatGPT usage — an NBER working paper based on a privacy-preserving, automated analysis of 1.5 million conversations drawn from a service with roughly 700 million weekly active users. The paper maps three years of adoption and shows AI use broadening beyond early adopters: gender gaps have largely closed (female-name share rising from 37% in Jan 2024 to 52% by July 2025), growth is fastest in low- and middle-income countries (adoption growth >4x that of high-income countries), and overall use is diversifying across demographics and geographies. Technically, the study categorizes interactions into Asking (49%), Doing (40%) and Expressing (11%), with Doing including drafting, planning and programming — writing is the most common work task while coding and self-expression remain niche. About 30% of consumer sessions are work-related and 70% non-work, yet both drive economic value by improving decision support, judgment, and productivity in knowledge-intensive roles — benefits that conventional GDP measures may miss. For the AI/ML community, the findings offer large-scale empirical signals about real-world utility, rising equity of access, key product use cases, and how usage cohorts deepen engagement as models improve.