OpenAI releases first-of-kind study revealing how people are using ChatGPT for everyday tasks (www.cnbc.com)

OpenAI and collaborators released a first-of-its-kind NBER working paper analyzing 1.5 million privacy-preserved consumer ChatGPT conversations to map who uses the service and for what. Key findings: non-work usage rose sharply to 73% of messages in June 2025 (from 53% a year earlier); three-quarters of conversations fell into "practical guidance, seeking information, and writing"; practical guidance (tutoring, how‑tos, ideation) was the single largest use; and about two‑thirds of writing requests were editing, critiquing or translating rather than generation from scratch. The paper also classifies interactions as Asking, Doing, or Expressing — roughly half of all messages were Asking (advice/information), while work-related messages were more often Doing (56%) as ChatGPT performed job tasks like writing. The study signals that large language models are becoming embedded in everyday life and acting more as advisors and productivity tools than outright job replacers, with ChatGPT reaching an estimated 10% of the world’s adults by July 2025. Demographics shifted toward gender parity and faster adoption in low- and middle‑income countries (growth >4x that of the richest countries), highlighting global reach and localization needs. Methodological caveat: the paper is an internal, non–peer-reviewed analysis authored by OpenAI’s economic research team and Harvard economist David Deming. Implications include prioritizing edit-and-assist features, workplace augmentation workflows, internationalization and safety/policy considerations as adoption broadens amid rising competition from Gemini, XAI and others.