Is generative AI displacing workers at economy-wide scale?

Brynjolfsson, Chandar, Chen and the Stanford Digital Economy Lab use high-frequency administrative payroll data from ADP, covering millions of U.S. workers through June 2026, to ask whether generative AI has displaced workers at scale. Their answer is "no evidence of widespread, economy-wide job displacement." The exception is narrow. Employment of young workers, ages 22 to 25, in AI-exposed occupations "now stands 19% below where it would be had it kept pace with that of their less-exposed peers," while "experienced workers show no comparable gap." The authors say this divergence "has widened steadily since we first documented it in August 2025."

The excerpt gives two mechanisms and a set of robustness checks. The divergence "operates primarily through reduced hiring of young workers rather than increased separations," so the adjustment happens at the point of entry rather than through layoffs. Likewise, "adjustment is occurring through employment rather than base compensation," so pay for incumbents is not what moves. The gap "persists when excluding technology firms and computer occupations, when controlling for exposure to interest-rate increases and for remote work, and across alternative measures of AI exposure." The authors also list the limits: the patterns "attenuate when controlling for education," "show some divergent trends predating generative AI," and are "more pronounced in the ADP analysis sample than in national survey benchmarks," with "some evidence of consistent patterns in government administrative data."

Set against the nearest notes, this excerpt narrows the picture they draw. The Upwork study ties ChatGPT's release to lower employment and earnings in highly affected occupations, but among freelancers on one platform Did ChatGPT's release reduce freelance writing work and pay?. The payroll data here locate the exposed-occupation shortfall in the hiring of one age group, with experienced workers unaffected. Read together, the two point to entry into the work as the channel to watch, though neither study measures the other's population. The Anthropic scenario model projects a falling labor share and stagnating knowledge-worker wages, which is a claim about pay Does AI growth inevitably shift wealth away from workers?. This excerpt reports early movement on the employment margin rather than the pay margin. The skill-demand note finds the same exposed-versus-less-exposed split in vacancy data, but as a divergence in skills rather than in who gets hired Is AI creating common skills across jobs or deepening divisions?. If the shortfall lands in junior hiring, the institutional argument in What makes accountable judgment scarce when AI cognition is cheap? becomes testable, since that note holds that outcomes depend on how expertise is learned. The excerpt measures the hiring gap, not that institutional cause.

The authors are explicit about what the excerpt does not establish. They call these "early, descriptive indicators—canaries in the coal mine—rather than causal estimates." The 19% is a counterfactual gap measured against less-exposed peers, and its attenuation under education controls means some of it may reflect who the young workers are rather than what their jobs expose them to. The excerpt also gives little detail on the exposure measures beyond saying alternative ones agree, and the ADP sample runs hotter than national survey benchmarks. The implication, at the strength the evidence allows, is that the youngest entry cohort in exposed occupations is a signal worth tracking, and the authors' public AI Economic Indicators are the means to do it. The excerpt does not show that generative AI caused the gap.