AIProof

Methodology

How we score AI exposure

AIProof's score comes from a real occupational dataset — the AI Work Index — not our imagination. It's a concrete number meant to turn a vague fear ("AI will take my job") into something you can inspect and argue with. It is not a prediction about you, and it isn't career advice.

What the number is

The score (0–100) is the role's AI task-exposure (exposure_v7): the share of its typical tasks that today's AI can already do or meaningfully assist with. The AI Work Index builds this by blending four independent sources — AIOE, the Anthropic Economic Index, Eloundou et al. (GPTs are GPTs), and the ILO — so no single model drives it. We map each role to its matching occupation and read the score straight from the data. Bands: AI-Proof, Resilient, Contested, Endangered, Critical.

Exposure is not the same as job loss. A high exposure score means AI can do a lot of the tasks — it does not mean the job disappears. Nursing scores high on task-exposure (charting, triage, admin) yet has very low actual job-loss risk, because demand is strong and the human parts are irreplaceable. Each role also carries a net job-loss risk figure (after market demand) in our data; the escape plan is built around widening that gap in your favour.

The four bottlenecks

Under the hood the exposure reflects how much a role leans on things AI still struggles with — the classic automation "bottlenecks":

Hands-on physical work in unpredictable settings (perception & manipulation) → lower exposure.
Social intelligence — care, trust, persuasion, accountability → lower exposure.
Novel, high-stakes judgment with real liability → lower exposure.
Routine language & data work → higher exposure. Note: unlike 2013-era studies, language creativity is now exposed too — models write and design well, so copywriting scores high, not low.

Honesty box

Scores for the vast majority of roles are read directly from the AI Work Index occupation table (we show the matched occupation, its confidence level, and the net job-loss figure on every plan). A small number of roles have no direct occupation match in the source (currently voice actor, teacher, firefighter) — those are clearly-labelled AIProof estimates, not dataset values. The single-country wage/employment fields in the source aren't used; only the exposure signal is.

Sources

← Back to AIProof