Data version 2026-11 · Maintained by the AIProofMe Team
Every occupation’s risk level comes from the U.S. Bureau of Labor Statistics (BLS), which rates how exposed each occupation is to AI. Three studies from OpenAI, Anthropic and Microsoft then decide where the job sits within that level. A score measures how much of a job AI can affect, not the chance that you will lose it.
| BLS AI exposure | Score range | Risk level |
|---|---|---|
| Very high | 75–95 | HIGH RISK |
| High | 55–74 | MEDIUM RISK |
| Moderate | 45–54 | LOW RISK |
| Low | 40–44 | LOW RISK |
In August 2026 the BLS published four categories of relative AI exposure (low, moderate, high, very high) for the 831 occupations it projects. To build them, BLS combined five external studies: three that score what AI could do in theory and two that measure how people actually use AI. It converted each study to a percentile rank, took the median of the theoretical and observed measures, and grouped occupations into the four categories.
We use the BLS category as published. It decides the score range in the table above, so a job BLS rates “very high” is always high risk on AIProofMe.
Inside each range, we place occupations using three of the studies BLS also relies on, with equal weight:
Each study is converted to a percentile rank across all occupations, the available ranks are averaged, and the job is placed within its range by that average: a higher average means a higher score within the same level.
Some occupations in our list are broad groups or “all other” categories that the studies do not score directly. For 129 of them we use the data of the closest occupations in the same Standard Occupational Classification (SOC) family. Each occupation page shows how many studies cover it.
Military occupations are not covered by BLS projections or by any of the three studies, so we do not give them a score. Their pages explain why and show official Department of Defense figures instead.
The free assessment starts from your occupation’s score. You pick the tasks you actually do from your occupation’s O*NET task list and mark the ones that take most of your time (they count double). We compare the AI exposure of your tasks, from OpenAI’s task-level ratings, with the average for all of your occupation’s tasks, multiply the difference by 80 and cap it at 15 points either way. The result stays within 40–95. Your age, how you work, your employer’s plans and your own AI use do not change the score; where research applies to them, the result page shows it separately with sources.
Task lists are based on O*NET 30.0 by the U.S. Department of Labor, Employment and Training Administration (USDOL/ETA), used under the CC BY 4.0 license. We shortened the task wording; USDOL/ETA has not approved, endorsed or tested these modifications.
The other two studies in the BLS rating (Felten, Raj & Seamans; Eisfeldt et al.) are included through the BLS category; we do not use them separately.
Download the data: BLS AI exposure categories (XLSX) · Microsoft Working with AI scores.
AIProofMe is built and maintained by the AIProofMe Team at Funovate Technology LLC. Questions or corrections: [email protected]. See also our editorial policy and legal notice.