What will AI actually change about your job? We break 48 common U.S. jobs into the tasks people really do, weight each task by how important it is, and show which ones AI can already do much of, which it can help with, and which stay human.
Exposure is not job loss. A job where AI helps with most tasks usually changes shape. The work that stays human often becomes the core of the role.
The four bands
- High: AI can already do much of 40% or more of the work.
- Transforming: AI touches 60% or more of the work, mostly as a helper.
- Moderate: AI touches 30% to 60% of the work.
- Low: AI touches less than 30% of the work.
How the model works
Break jobs into tasks. Each job’s task list comes from O*NET, the U.S. Department of Labor’s occupation database, with each task weighted by its importance rating from people who do the job.
Rate each task. Each task carries a published rating from trained human reviewers: a large language model can cut the time it takes by half (AI can do much of it), it can only with extra software built on top (AI can help), or it cannot (stays human).
Roll it up. The three shares are importance-weighted percentages of the job’s tasks, and the band follows from them.
These ratings were last reviewed in October 2026. TrustNoRobot re-rated every task against generally available AI tools using AI-assisted review, replacing the original 2023 research ratings (Eloundou et al., “GPTs are GPTs”), so this update moved more than future quarterly reviews should. Task lists and importance weights come from the O*NET 31.0 database. Jobs also differ by workplace: this describes the typical U.S. version of each job, not any one person’s role.
Sources: O*NET 31.0 Database (August 2026), U.S. Department of Labor. Task exposure ratings: last reviewed October 2026 by TrustNoRobot using AI-assisted review against current AI tools, originally based on Eloundou, Manning, Mishkin and Rock, “GPTs are GPTs” (2023). Employment: U.S. Bureau of Labor Statistics, May 2021.