OpenAI Data Show Work-Related ChatGPT Messages Were Less Concentrated in Top Activities

OpenAI released data Oct. 26 mapping work-related ChatGPT messages to 165 activity categories across nearly 100 countries. In a comparison of 74 countries with data for every quarter, the 10 most common activities in each country accounted for a smaller average share of classified work messages in Q3 2026 than in Q1 2025.

Those activities represented an average 51.1% of classified work messages in Q3 2026, down from 56.0% in Q1 2025. The 4.9 percentage-point decline indicates that messages were less concentrated in the most common categories. It does not explain why the distribution changed.

The quarterly series runs from Q1 2025 through Q3 2026. Its categories, called Intermediate Work Activities, include preparing instructional materials, developing marketing content, programming and analyzing business data. They use O*NET, a framework that groups occupation-specific tasks into broader activities shared across occupations.

The wider Signals v3.0 dataset covers July 2024 through September 2026 and classifies one million consumer ChatGPT messages per month from accounts with a declared age of at least 18. It excludes Enterprise messages, Codex use, accounts that were deleted and users who opted out of allowing their data to be used for training. Country-level activity shares require an effective sample of at least 1,000 messages per country-quarter.

Automated classifiers were applied to de-identified data, and researchers did not view raw messages. Published statistics are aggregated and include differential-privacy noise.

The figures cover classified messages on individual ChatGPT plans. They describe how those messages were distributed among work activities; they do not by themselves establish workforce-wide AI adoption, productivity gains or job losses.