OpenAI has released a new dataset through OpenAI Signals that maps consumer ChatGPT use to 165 standardized work activities, called Intermediate Work Activities (IWAs), across nearly 100 countries. The data is quarterly, covering the first quarter of 2025 through the third quarter of 2026, and is drawn from work-related messages on individual ChatGPT plans. By connecting messages to the O*NET framework, the release lets researchers compare ChatGPT usage with occupational and employment data, track changes over time, and pair AI usage with other economic indicators.

The data reveals that work-related ChatGPT use has become less concentrated over the observed period. Across 74 countries present in every quarter, the ten most common activities represented an average of 51.1% of each country’s classified work messages in Q3 2026, down from 56.0% in Q1 2025. That shift points to countries specializing in different types of AI-assisted work over time, potentially reflecting differences in the kinds of work performed or varying access to AI tools.

The release also includes paired visualizations comparing ChatGPT usage with the activity mix of U.S. jobs. The two distributions differ substantially, largely because about 43% of the employment-weighted task mix for mapped U.S. occupations falls into “Other IWAs” — categories such as moving materials, operating vehicles, and maintaining health or medical records that did not meet the threshold for consistent representation in the original U.S. ChatGPT sample.

OpenAI notes that this data covers individual ChatGPT plans and excludes enterprise messages and Codex use. To protect privacy, researchers do not read individual conversations; automated systems classify messages after identifying information is removed, and the published aggregate statistics are protected by suppressing small groups and adding statistical noise.