Skip to content
After AI

Bibliography

Every number, and where it came from.

23 sources. Nothing on this site is cited that is not listed here, and every source listed here is cited somewhere. The build fails if that stops being true.

How to read a source

Three kinds of claim appear on this site, and they are labelled differently: measured (payroll data, official projections, task scoring), forecast (employer surveys and employer expectations, always marked), and editorial (the 0–3 scoring on each job profile, and the framing). Confusing the first two is the most common error in this field.

Where a source has a commercial interest — a governance vendor, a job board counting its own postings — the note says so, and the number is presented as directional.

Coverage gaps

The evidence base skews Anglophone and to high-income countries: the employment-effect study is US payroll data, the exposure index is global but calibrated on one national task taxonomy, and the advertisement analysis covers 27 countries.

Low- and middle-income labour markets are thinner here, and informal work barely appears at all — which matters, because that is where most of the world's employment is.

Peer-reviewed & working papers

Task-level measurement and observed employment effects.

3

Employer & advertisement surveys

Large-scale expectations and hiring data. Forecasts, not outcomes.

2

Official statistics

Government projections and wage data.

5

Institutional & vendor reports

Useful, but read the interest behind them — noted per source.

11

Journalism

Where the ranked lists and primary quotes are reported.

2
Corrections

If a figure here is wrong, the fix is a data file.

Every chart series lives in one typed file with its source id attached, and every job profile is a folder holding a validated data file and a written note. A number cannot reach a chart without a reference, and a reference cannot exist without an entry on this page.