OpenAI releases ChatGPT work-use data by country
OpenAI has published quarterly data that maps how consumers in nearly 100 countries use ChatGPT for work onto 165 standard O*NET work activities, so economists can compare AI use with labor data.
OpenAI announced a new release of its OpenAI Signals data. It sorts work-related ChatGPT messages into 165 Intermediate Work Activities (IWAs), the task categories that O*NET uses across occupations. The data covers nearly 100 countries and runs quarterly from Q1 2025 through Q3 2026.
O*NET is a standardized framework, so researchers can link ChatGPT usage to other occupational and employment data. They can also compare countries and follow trends over time. Activity categories range from preparing instructional materials and developing marketing content to programming and analyzing business data.
OpenAI reported two early findings:
- Countries became more specialized. Over time, work use spread out from the most common activities, and different countries moved toward different types of work. OpenAI says this could reflect differences in the work done in each country or in access to AI tools. It does not claim a cause.
- ChatGPT use looks very different from the U.S. economy. About 43% of the employment-weighted task mix for mapped U.S. occupations falls into an "Other IWAs" group, which includes activities like moving materials, operating vehicles and maintaining medical records. These activities did not show up consistently enough in OpenAI's U.S. ChatGPT sample to be reported individually.
The dataset has limits. It covers only individual ChatGPT plans and leaves out enterprise messages and Codex use. OpenAI says messages are classified by automated systems after identifying information is removed. Results for small groups are withheld, and statistical noise is added. The announcement does not say how accurate the classifier is.