The rapid ascension of Large Language Models (LLMs) and conversational AI has undeniably reshaped our technological landscape. Yet, for all the buzz and speculation, a critical question has remained largely unanswered: how are people actually using these powerful tools in their daily lives and work? Until now, much of our understanding has relied on self-reported surveys or analyses of smaller, specialized datasets. A groundbreaking NBER Working Paper, “How People Use ChatGPT”, authored by a team including researchers from Duke University, Harvard University, and OpenAI, finally pulls back the curtain on this phenomenon. This comprehensive study, leveraging internal ChatGPT message data through a novel privacy-preserving methodology, offers an unprecedented look at how the world’s largest consumer chatbot is being adopted and utilized. With ChatGPT reaching over 700 million weekly active users and handling 18 billion messages per week by July 2025, these findings are not just interesting; they are foundational for anyone building, investing in, or strategizing around AI. The Evolving Landscape of ChatGPT Usage: Beyond Work Productivity The Evolving Landscape of ChatGPT Usage: Beyond Work Productivity The paper’s core findings paint a dynamic picture, challenging some prevailing narratives and spotlighting unexpected trends. The Rise of Non-Work Engagement: Perhaps the most striking revelation is the dramatic shift towards personal, non-work-related use. While AI’s economic value has often been framed through the lens of enterprise productivity, this study reveals that non-work messages have grown significantly faster than work-related ones, expanding from 53% to over 70% of all consumer ChatGPT usage between June 2024 and June 2025. This suggests an enormous impact on “home production” and a substantial consumer surplus, estimated at a staggering $97 billion in the US alone for 2024. This indicates AI is becoming deeply integrated into our personal lives, enhancing everything from learning to leisure.
The Three Pillars of Conversation Topics: When it comes to the content of conversations, three broad categories dominate: “Practical Guidance,” “Seeking Information,” and “Writing.” Collectively, these account for nearly 78% of all ChatGPT interactions. The Rise of Non-Work Engagement: Perhaps the most striking revelation is the dramatic shift towards personal, non-work-related use. While AI’s economic value has often been framed through the lens of enterprise productivity, this study reveals that non-work messages have grown significantly faster than work-related ones, expanding from 53% to over 70% of all consumer ChatGPT usage between June 2024 and June 2025. This suggests an enormous impact on “home production” and a substantial consumer surplus, estimated at a staggering $97 billion in the US alone for 2024. This indicates AI is becoming deeply integrated into our personal lives, enhancing everything from learning to leisure. The Three Pillars of Conversation Topics: When it comes to the content of conversations, three broad categories dominate: “Practical Guidance,” “Seeking Information,” and “Writing.” Collectively, these account for nearly 78% of all ChatGPT interactions. Writing’s Nuance: Writing is, by far, the most common work-related use case, making up 40% of work messages in June 2025. However, the study provides a critical insight: about two-thirds of these writing tasks involve modifying existing text (editing, critiquing, summarizing, translating) rather than generating entirely new content from scratch. This highlights AI’s role not just as a creator but as a powerful editor and enhancer of human-generated content; a true augmentation tool. Guidance and Information Seeking: “Practical Guidance” and “Seeking Information” underscore ChatGPT’s role as a hyper-personalized search engine and tutor. Notably, approximately 10% of all messages are requests for tutoring or teaching, pointing to AI’s significant potential in education. The Ascendance of “Asking”: AI as a Strategic Co-Pilot To delve deeper into user intent, the researchers introduced a novel classification system: “Asking, Doing, or Expressing”. “Asking” messages involve seeking information or advice to inform decisions. “Doing” messages request ChatGPT to perform tasks that produce output, like drafting an email or writing code. “Expressing” messages are for conveying views or feelings without seeking information or action. The data reveals a compelling trend: as of July 2025, 49% of user messages are “Asking,” while 40% are “Doing,” and 11% are “Expressing.” Importantly, “Asking” messages have grown faster than “Doing” messages over the past year and are consistently rated as having higher quality by users. This finding reinforces the idea that ChatGPT provides significant economic value through decision support, particularly in knowledge-intensive jobs. Rather than merely acting as a “co-worker” that produces output, AI is increasingly functioning as a “co-pilot” that gives advice and enhances human problem-solving capabilities. This is further supported by the analysis of O*NET Work Activities, which found that 81% of work-related messages align with “obtaining, documenting, and interpreting information” and “making decisions, giving advice, solving problems, and thinking creatively”. These activities were consistently high across various occupations, from management to STEM. Demographic Shifts: Broadening Access and Impact Demographic Shifts: Broadening Access and Impact The study also shed light on who is using ChatGPT and how those demographics are evolving: Gender Gap Narrows Dramatically: Early adopters were disproportionately male. However, by June 2025, the gender gap had narrowed considerably, potentially closing completely, with active users becoming slightly more likely to have typically feminine first names. Users with typically female first names showed a relative inclination towards “Writing” and “Practical Guidance,” while those with typically male first names leaned towards “Technical Help,” “Seeking Information,” and “Multimedia”. Global and Youth-Driven Growth: ChatGPT usage has grown relatively faster in low- and middle-income countries over the last year. Furthermore, nearly half of all messages sent by adults came from users under the age of 26, although age gaps have shown some narrowing recently. Education and Occupation Matter for Work Usage: Educated users and those in highly-paid professional occupations are significantly more likely to use ChatGPT for work-related messages and, within that context, for “Asking” rather than “Doing” tasks. This further solidifies the notion of AI as a decision-support tool for complex, knowledge-intensive roles. Implications for the AI/Tech Industry These revelations carry profound implications for how we design, develop, and deploy AI: Refined Product Strategy: Prioritize “Co-Pilot” over “Co-Worker.” The clear trend towards “Asking” and the higher quality ratings for such interactions signal a demand for AI that excels as a strategic advisor, intelligent research assistant, and nuanced editor. Future product roadmaps should lean into sophisticated decision support systems, highly customizable guidance, and advanced writing assistance tools that go beyond mere content generation, focusing instead on refining and enhancing user inputs. Unlocking New Markets: The Power of “Home Production.” The massive and accelerating non-work usage reveals an enormous untapped market for AI solutions that enrich personal lives, manage households, facilitate learning, and support well-being. This expands the perceived economic value of AI far beyond enterprise-level productivity gains. Rethinking AI’s Role in Knowledge Work: The emphasis on decision support suggests AI is an intelligence augmenter rather than solely a labor automator. For professionals, AI is becoming an indispensable tool for better problem-solving and creative thinking. Companies should frame AI solutions around empowering human intellect, not just replacing tasks, and measure ROI accordingly. Setting New Standards for Responsible AI Research: The paper’s privacy-preserving methodology aka utilizing automated LLM-based classifiers on de-identified and PII-scrubbed data establishes a crucial precedent. It demonstrates that deep, granular insights into user behavior can be achieved without compromising individual privacy, a critical concern as AI integration deepens. Dispelling Myths with Data: The study’s findings on the relatively small shares of usage dedicated to computer programming (4.2%) and companionship/social-emotional issues (1.9%) for ChatGPT directly contrast with some other industry analyses or popular perceptions. This underscores the importance of large-scale, direct usage data over self-reports or niche studies to gain an accurate understanding of mass-market AI adoption. It suggests that while these uses exist, they are not the dominant ways the general public engages with ChatGPT. Conclusion The “How People Use ChatGPT” paper offers an indispensable data-driven perspective on the real-world impact of generative AI. It reveals a technology deeply embedded in personal enrichment and evolving into a sophisticated decision-support tool, especially in knowledge-intensive professions. The demographic shifts indicate a broadening of AI’s reach across genders and income levels globally, hinting at a more inclusive technological future. These insights compel us to reconsider our assumptions about AI’s ultimate purpose and potential. Are we ready to fully embrace AI not just as a doer of tasks, but as a crucial partner in thinking, learning, and navigating our increasingly complex world?

The is an opinion piece based on the author’s POV and does not necessarily reflect the views of HackerNoon.

This story contains AI-generated text. The author has used AI either for research, to generate outlines, or write the text itself. 

How People Use ChatGPT

About Author

Comments

TOPICS

THIS ARTICLE WAS FEATURED IN

Related Stories

Untitled Story

3 Surprising Ways AI is Redefining the Search for Cures to Rare Diseases

I Built a WordPress-like Hook System for Node.js to Decouple My Code

Ethical AI Governance Could Be Every Startup’s Secret Weapon

Avoiding Configuration Bleed: Cleanly Separating Staging and Production on Netlify (in a Monorepo)

The Noonification: How Often Do NFTs Pass The Howey Test? (1/13/2023)

3 Surprising Ways AI is Redefining the Search for Cures to Rare Diseases

I Built a WordPress-like Hook System for Node.js to Decouple My Code

Ethical AI Governance Could Be Every Startup’s Secret Weapon

Avoiding Configuration Bleed: Cleanly Separating Staging and Production on Netlify (in a Monorepo)

The Noonification: How Often Do NFTs Pass The Howey Test? (1/13/2023)

Light-Mode

Classic

Newspaper

Dark-Mode

Neon Noir

Minty

HN StartUps