Stanford payroll data shows AI's hiring freeze on young workers widening, not stabilising
The employment gap for 22-to-25-year-olds in AI-exposed jobs grew from 15% to 19% in a year, and the mechanism is reduced hiring, not layoffs.
Original cover art, generated for this story. THE VISSION does not republish third-party press imagery.
- Stanford's Digital Economy Lab, using ADP payroll data, finds employment among workers aged 22–25 in highly AI-exposed occupations is 19% below where it would be had it kept pace with less-exposed peers — up from a 15% gap a year earlier.
- The researchers find no widespread, economy-wide displacement; the effect is concentrated specifically in young, inexperienced workers.
- The adjustment operates through reduced hiring of young workers rather than increased separations of those already employed.
- Employment fell in occupations built on codified knowledge — the kind textbooks and documentation can teach — while employment among experienced workers in tacit-knowledge occupations held flat or rose.
New payroll data from Stanford's Digital Economy Lab puts a specific, widening number on a question that has mostly been argued in the abstract: how much is AI actually displacing entry-level work. Researchers Erik Brynjolfsson, Bharat Chandar and Ruyu Chen, using ADP payroll records and occupational AI-exposure measures, find that employment among workers aged 22 to 25 in highly AI-exposed occupations now sits 19% below where it would be had it tracked employment among similarly aged workers in less-exposed occupations. A year earlier, in July 2025, that gap was 15%.
The study is explicit about what it is not finding. "We do not see widespread, economy-wide job displacement associated with AI," the authors write — average employment across their full ADP sample rose roughly 6% from November 2022 to June 2026, while employment in the most AI-exposed occupations grew only about 4%. The effect is narrow and concentrated, not a broad contraction.
Where it concentrates is the more useful finding. The adjustment shows up almost entirely as reduced hiring of young workers rather than as increased separations among people already in a job — companies are not firing junior staff so much as declining to hire replacements for the tasks AI can now do. And the divide tracks a specific distinction: employment fell in occupations built on codified knowledge, the kind that can be taught through documentation and formal training, while employment among experienced workers in occupations built on tacit knowledge — skill acquired through practice and mentorship rather than text — held flat or rose. "This pattern is consistent with a world in which generative AI is particularly effective at reproducing and applying knowledge that has already been encoded in text," the authors write.
The finding updates the same team's earlier "Canaries in the Coal Mine?" research, and the direction of travel is the story: a real, measurable gap that widened over the past year rather than stabilising, even as the study's own headline finding is that there is no broad, economy-wide AI jobs crisis.
For anyone hiring, this reframes the AI-jobs question from "how many jobs will AI take" to "who bears the adjustment cost" — and the answer, in this data, is specifically people without the tacit knowledge that comes from having already had the job. A company that stops hiring juniors because AI now covers the codified parts of entry-level work is making a rational near-term choice that removes its own pipeline for the tacit knowledge those juniors would have built into senior staff a decade from now. That is a cost the quarterly hiring decision does not price in.