How the World Uses ChatGPT: Non‑Work Growth, Decision Support, and Writing at Work

Using privacy‑preserving classifiers over 1.1M sampled messages, the authors show ChatGPT use skews toward Practical Guidance, Seeking Information, and Writing, with Writing dominating work tasks. Non‑work usage has grown faster than work usage, and Asking now exceeds Doing overall, while Doing (mostly Writing) remains largest at work. Adoption has diffused broadly across demographics and countries, and the main economic value arises from decision support and communication rather than task automation.

Key Points

• Non‑work use has grown faster than work use, rising from 53% to 73% of consumer messages between mid‑2024 and mid‑2025.
• Three topics—Practical Guidance, Seeking Information, and Writing—dominate usage; Writing is the top work task and mainly involves improving user‑provided text.
• Intent analysis shows Asking (information/advice) now exceeds Doing (content production); at work, Doing predominates and is mostly Writing.
• Across occupations, ChatGPT is used for information gathering, documentation/interpretation, and decision support; coding comprises only ~4% of messages.
• Adoption has broadened demographically: the early male skew has closed, use is heavy among under‑26s, and growth is fastest in low‑ and middle‑income countries.

Sentiment

Mixed but leaning skeptical: most comments restate or accept the paper’s findings, while the main opinionated comment is pessimistic about monetization due to the consumer-dominated usage.

In Agreement

• Non-work queries now dominate usage (around 70–73%), reflecting a strong consumer tilt.
• Most usage clusters into practical guidance, information seeking, and writing (~80%).
• At work, writing is the main task and mostly involves editing or improving user-provided text.
• Rapid, global adoption with fastest growth in lower-income countries and a user base trending younger and more gender-balanced.
• Overall interaction quality is improving, with higher satisfaction for Asking-type interactions.

Opposed

• The consumer-heavy usage mix bodes poorly for long-term economics because B2C users are harder to monetize.
• Corporate pushes toward professional use appear less impactful, prompting pivots to find new use cases.
• Skepticism that the trend will end well given the monetization challenges implied by the data.