21 months of aggregate ChatGPT usage data (July 2024 – March 2026). OpenAI calls it “adoption.” Through the Discontinuity lens, it’s the most comprehensive displacement dataset ever published by an AI lab.
What people use ChatGPT for, month by month.
Writing dominates at ~24%. Technical Help and Practical Guidance together account for massive shares — professional work tasks outsourced to AI. Self-expression growing fastest = creative work consumed too.
Share of messages classified as work-related vs personal.
Work share declined 51% → 31%. Copers cite this. But total volume grew enormously. 31% of a vastly larger pie = MORE absolute work displacement. The declining share masks accelerating replacement.
“Doing” = ChatGPT performing actual work tasks. Not researching. Not brainstorming. Doing. At 46%, nearly half of work interactions are direct task execution. A replacement worker measured in real-time.
25-34 year olds dominate (~40%+). Early-career professionals outsourcing their work to AI. Building AI-dependency instead of skills.
Which specific human work activities ChatGPT is replacing, by average share of work messages.
| Rank | IWA Code | Work Activity | Avg Share |
|---|---|---|---|
| 1 | 4.A.2.b.1.I07 | Processing Transactions | |
| 2 | 4.A.3.b.6.I12 | 4.A.3.b.6.I12 | |
| 3 | 4.A.3.b.6.I16 | 4.A.3.b.6.I16 | |
| 4 | 4.A.3.b.1.I01 | Performing for or Working Directly with the Public | |
| 5 | 4.A.2.b.2.I12 | 4.A.2.b.2.I12 | |
| 6 | 4.A.3.b.1.I04 | 4.A.3.b.1.I04 | |
| 7 | 4.A.1.a.1.I04 | Getting Information | |
| 8 | 4.A.4.a.1.I01 | Processing Information | |
| 9 | 4.A.2.b.2.I18 | 4.A.2.b.2.I18 | |
| 10 | 4.A.4.a.1.I03 | Interpreting Data/Results | |
| 11 | 4.A.4.a.1.I02 | Analyzing Data/Information | |
| 12 | Other IWA | Other IWA | |
| 13 | 4.A.4.a.2.I03 | Testing/Debugging Software | |
| 14 | 4.A.1.a.1.I07 | Judging the Qualities of Objects, Services, or People | |
| 15 | 4.A.4.b.6.I07 | 4.A.4.b.6.I07 | |
| 16 | 4.A.1.b.3.I03 | Estimating Time | |
| 17 | 4.A.2.a.4.I04 | Planning Events/Programs | |
| 18 | 4.A.3.b.6.I13 | 4.A.3.b.6.I13 | |
| 19 | 4.A.2.a.1.I07 | Problem Analysis | |
| 20 | 4.A.1.a.1.I18 | Compiling Data/Statistics |
Top codes cluster around information processing, creative work, computer interaction, written communication. Core activities of knowledge workers: analysts, writers, programmers, designers, marketers, researchers.
Plus subscribers use ChatGPT for work at significantly higher rates. Professionals paying for enhanced AI are using it to replace human work output — not for casual chat.