At a glance
- Source
- Brynjolfsson, Chandar and Chen (2025). Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence. Stanford Digital Economy Lab working paper (revised 12 August 2026). digitaleconomy.stanford.edu · Not peer-reviewed: other researchers have not yet checked it for a journal.
- How it was done
- The authors used monthly payroll records from ADP, a large US payroll company, and compared employment by age group in occupations more and less exposed to AI. Data analysis. It builds a measure from existing data, such as job and task records.
- Who took part
- Payroll records for millions of US workers, up to June 2026.
- Where
- United States.
“It operates primarily through reduced hiring of young workers rather than increased separations.”
What it found
An occupation counts as exposed when AI can do many of its tasks.
- No sign of job losses across the whole economy. The authors find no evidence of widespread displacement.
- Young workers in exposed jobs. Employment of 22–25-year-olds in AI-exposed occupations is 19% below where it would be had it kept pace with less exposed peers. Experienced workers show no such gap.
- A widening gap. The gap has grown steadily since the authors first reported it in August 2025.
- Fewer hires, not more departures. The gap comes mainly from fewer young people being hired, not from more people leaving.
- Where AI replaces tasks. The fall is concentrated in occupations where people mostly use AI to do tasks instead of them. Where people mostly use AI to help them, employment is flat or rising, especially for experienced workers.
- Jobs, not pay. Firms are adjusting the number of people they employ, not base pay.
When the result holds
The figure has changed between versions of the paper. The 19% comes from the revision of 12 August 2026, with data up to June 2026. Earlier versions reported smaller gaps. Cite the version and its date.
The study measures who gets hired. Studies that measure people already in work, or the economy as a whole, find much smaller effects or none so far. Whether AI is cutting entry-level hiring is disputed.
Limits
- The study describes what goes together. It does not show that AI caused the gap. Other changes, such as interest rates or cuts in the technology sector, could line up with exposure.
- ADP’s records do not cover the whole US economy in proportion. The gap is larger in ADP data than in national surveys, and it narrows once workers’ education is taken into account.
- An occupation’s exposure score is not the same as a firm actually using AI.
- United States only. A working paper, not peer-reviewed.
Practices that use this study
This study supports:
This study limits or contradicts: