OpenAI’s Economic Research group released a dataset called OpenAI Signals that links ChatGPT usage to 165 Intermediate Work Activities (IWAs), providing a quarterly view of how people use the model for work tasks. The release covers the period from Q1 2025 through Q3 2026 and includes activity data across nearly 100 countries.
What the data shows
The dataset maps ChatGPT messages to O*NET-derived IWAs, which group occupation-specific tasks into broader activity categories. OpenAI reports wide variation across countries in the mix of work-related messages. Across 74 countries observed in every quarter, the ten most common activities accounted for an average of 51.1% of each country’s classified work messages in Q3 2026, down from 56.0% in Q1 2025, a change OpenAI attributes to increasing specialization of activity mixes across countries.
OpenAI also compared ChatGPT usage to the activities performed in U.S. occupations. The organization notes a substantial difference between the distributions: roughly 43% of the employment-weighted task mix for mapped U.S. occupations falls into “Other IWAs”—tasks such as moving materials, operating vehicles, and maintaining health records—which did not meet the threshold for consistent representation in the original U.S. ChatGPT sample.
Scope, methods and privacy
The release covers work-related messages on individual ChatGPT plans and explicitly excludes enterprise messages and Codex use. OpenAI describes an automated classification pipeline that operates after identifying information is removed. The published statistics are aggregated; results for small groups are withheld and statistical noise is added to protect individual messages.
OpenAI invites researchers to explore and download the data to study differences across countries, track changes over time, and link ChatGPT use to other economic indicators. The dataset is presented quarterly and is dated October 26, 2026.
Original source: OpenAI News