How it works

Every number comes from published data and a transparent model. A language model never sets a score; it only explains.

From a job to a sourced score

My job

›

O*NET tasks

›

AI + robotics exposure

›

Timeline band

›

Sourced range

AI exposure from Eloundou et al. (2023); robotics from O*NET work context; the timeline shifts for local demand (BLS OEWS), union coverage, and age.

Does it hold up?

Average 5-year employment change by our risk band (702 occupations). The highest-risk band actually lost jobs.

Individual outcomes are predicted weakly (cross-validated R-squared about 0.046), which is why results are shown as ranges, not verdicts. The 2020 base year spans COVID, so this validates the ranking, not a precise score.

What we never use

Gender, race, disability, and other protected characteristics are never inputs to the score. Age is used only to work out your remaining working years.

O*NETEloundou et al. (2023), GPTs are GPTsO*NET Work ContextBLS CPI-UBLS OEWS

Data snapshot SNAPSHOT-2026-11-06.