OpenAI releases Signals data on ChatGPT work usage across 165 activities and 100 countries

• OpenAI Economic Research released Signals v3.0 data covering 165 work activities across nearly 100 countries, spanning Q1 2025 through Q3 2026.
• Work-related usage became less concentrated over time; across 74 countries observed in every quarter, the ten most common activities represented 51.1% of classified work messages in Q3 2026, down from 56.0% in Q1 2025.
• The data maps ChatGPT usage to Intermediate Work Activities (IWAs) based on the O*NET framework, allowing researchers to compare AI usage with occupational and employment data.
• The dataset excludes enterprise messages and Codex use, focusing on individual ChatGPT plans, with privacy protections including automated classification and statistical noise.

Data Scope and Methodology

• Covers work-related messages on individual ChatGPT plans, excluding enterprise and Codex usage.
• Uses O*NET to group occupation-specific tasks into broader Intermediate Work Activity (IWA) categories.
• Privacy measures include removing identifying information, withholding results for small groups, and adding statistical noise.

Research Applications

• Enables comparison of ChatGPT usage patterns with actual work performed in the U.S. economy.
• Highlights differences between AI usage and employment-weighted task mixes, with about 43% of U.S. tasks falling into "Other IWAs" not individually reported in the ChatGPT sample.