Data
The numbers,
open.
One table with every job we score. Free to use under the CC BY 4.0 licence, with a link back.
What each column means
- job: The job title, as O*NET lists it.
- onet_code: The O*NET code for the job, for example 29-1141.00.
- sector: The broad group the job belongs to.
- ai_score: Our score, 0 to 100. The average of the study positions that cover the job, rounded.
- openai_position: Where the job ranks among all jobs in the OpenAI study, 0 to 100. Blank if the study does not cover it.
- anthropic_position: The same for the Anthropic study.
- microsoft_position: The same for the Microsoft study.
- us_average_pay_usd: Average yearly pay in the United States, in dollars, not adjusted for inflation.
- us_jobs: How many people do the job in the United States.
- pay_and_jobs_year: The survey year the pay and job figures are from.
- jobs_change_since_2021_pct: Change in the number of US jobs since 2021, in percent. Blank when the figures are too thin to trust.
- page: The job's page on this site.
Where each figure comes from
- OpenAI: Could AI help do each task in half the time? Judged by people, 2023. Source files
- Anthropic: How much people really use AI for the tasks in this job. Measured from use, 2025. Source files
- Microsoft: How well AI handled this kind of work in real Copilot chats, 2024. Source files
- Pay and jobs: US Bureau of Labor Statistics, Occupational Employment and Wage Statistics. Source
- Jobs and tasks: O*NET, from the US Department of Labor. Source
How the score is worked out
Each study is turned into the job's position among all jobs in that study, from 0 to 100. The score is the average of those positions. 853 of the 879 jobs are covered by all three studies. The full method.
Things to know before you use it
- The three studies measure different things, so their positions can be far apart for one job. Use the three columns, not only the average, when that matters.
- Several jobs can share one government pay code. Count each code once when you add up workers or average pay.
- A change in jobs or pay is not proof that AI caused it.
- All figures are for the United States.
Change log
- 2026-11: scores rebuilt from the source files.
This version was built in November 2026.