This note will explain, in plain language, where each exposure number on this site comes from. The short version: the occupation ranking combines three published academic measures — Felten's AI Occupational Exposure index, Eloundou et al.'s "GPTs are GPTs" ratings, and Eisfeldt et al.'s postings-based measure — alongside an observed-usage measure built from Anthropic's Economic Index: real people using AI on real tasks, matched to the official O*NET task list. The theoretical measures estimate how exposed a job could be given its tasks; the observed measure records what people actually use AI for.
The point worth stating up front, because it shapes everything: these measures are shown side by side and never blended into one headline number. They often disagree, and on this site the disagreement is treated as information, not noise.
The complete documentation — formulas, weights, validation checks, and the honest list of what each score can't tell you — already lives at How scores are computed. This post will be the reader-friendly walk through the same ground, and it will add nothing the documentation doesn't already state. The underlying sources are catalogued on Data Sources.