A new study by OpenAI’s Economic Research team, in collaboration with Duke and Harvard, offers the most comprehensive look yet at how people actually use ChatGPT, based on a privacy-preserving analysis of over 1.1 million real user conversations from May 2024 to June 2025. The findings reveal that ChatGPT is being used far less as a productivity tool and far more as a thinking partner. The study shows that 49% of messages are “asking” — seeking advice, brainstorming ideas, or exploring options — while only 40% are “doing,” such as drafting emails or generating content, and 11% are “expressing” emotions or opinions. Notably, “asking” is growing faster than “doing,” suggesting users value ChatGPT for cognitive support over output generation. The data also shows a dramatic shift in user demographics. Women now make up 52% of ChatGPT users with identifiable first names, a major turnaround from 2022 when men dominated the user base. Women are more likely to use the tool for writing and practical guidance, while men lean toward technical help and information-seeking. Young users, particularly those aged 18 to 25, account for nearly half of all usage, with 46% of messages coming from this group. However, they are less likely to use ChatGPT for work — only 22.5% of their messages are work-related, compared to 31.4% among users aged 36–45, the highest rate across all age groups. Non-work use has surged from 53% in June 2024 to 73% by June 2025, indicating that ChatGPT has become a central tool for personal learning, planning, and creative exploration. The most common use cases are practical guidance (e.g., tutoring, how-to advice), seeking information (like traditional search), and writing — with two-thirds of writing requests involving editing, summarizing, or translating existing text, not creating from scratch. Surprisingly, coding only accounts for 4.2% of all messages, challenging the myth that ChatGPT is primarily a developer’s tool. The study also found that “asking” messages are rated higher in quality and user satisfaction than “doing” messages, reinforcing the idea that engaging in dialogue with AI leads to better outcomes. This aligns with personal experiences of users who report deeper thinking and better decisions when they frame prompts around exploration and critique rather than direct commands. Despite the scale of use — over 700 million weekly users sending nearly 18 billion messages per week — the study highlights that AI adoption in the workplace remains underwhelming. Only 27% of interactions are work-related, suggesting many organizations have yet to unlock real value from AI tools. Meanwhile, personal use continues to grow, with users treating ChatGPT as a digital sounding board for ideas, decisions, and self-reflection. Privacy was a priority: no human reviewed messages. Instead, AI classifiers were used to categorize conversations while scrubbing personal data. The study’s findings, though not peer-reviewed, offer a rare, data-driven window into everyday AI behavior. The research underscores a key insight: the real power of ChatGPT lies not in replacing human work, but in enhancing human thinking. The most valuable interactions happen when users ask better questions, challenge assumptions, and refine ideas — not just delegate tasks. For individuals and teams, the takeaway is clear: design your AI sessions around inquiry, not output. Use prompts that invite debate, propose criteria first, and turn answers into action plans. In doing so, you tap into the true potential of AI as a collaborator in the creative and decision-making process. Ultimately, the study reveals that ChatGPT is not just a writing assistant or coding tool — it’s becoming a cognitive partner for millions, helping them think more clearly, act more intentionally, and grow more creatively.