Stanford Finds a 19% Employment Gap for Young Workers in AI-Exposed Jobs

Workers ages 22 to 25 in the most AI-exposed occupations now have employment levels 19% below peers in less-exposed fields, according to updated research from Stanford economists. The gap has widened even as relative employment differences across the full economy remain small.

The researchers separate exposure from use

The August update of Canaries in the Coal Mine uses anonymized, high-frequency payroll data aggregated by ADP. It ranks occupations with an existing labor-market-impact gauge and the Anthropic Economic Index, which examines how occupations use Claude in everyday work.

Two types of AI use

• Automation-oriented use means AI replaces work previously done by people. Occupations with more of this use had the weakest relative entry-level employment.
• Augmentation-oriented use means AI helps workers perform tasks they still do. Entry-level employment outcomes in those occupations were more mixed.

The researchers say the pattern is consistent with automation-oriented AI substituting for labor, while complementary uses are associated with flat or rising employment. That is an observed relationship, not proof that AI use caused each employment change.

The young-worker gap widened

Employment gap for workers ages 22 to 25

13%→19%

percent

The reported employment gap between young workers in the most AI-exposed occupations and peers in less-exposed occupations rose from 13% to 19%.

Hiring, rather than exits, drives the pattern

Since 2022, employment for 22- to 25-year-olds fell about 11% in the 40% of occupations judged most AI-impacted. It grew 10% in the 60% judged least impacted. The difference is mainly associated with lower hiring rates, rather than increased firings or quitting, and appears primarily in employment levels rather than pay rates.

That distinction limits the finding. The study found little to no relative employment difference economy-wide between the occupations judged most and least exposed to AI; its clearest effect is among workers ages 22 to 25.

Formal knowledge shows a second divide

The researchers used required formal education in O*NET as a proxy for codified knowledge: knowledge taught through schooling, documents, or procedures. Occupations with more codified knowledge had slower entry-level employment growth. Jobs relying more on tacit knowledge, built through experience, showed faster growth for mid-career and senior workers.

Occupations with more college graduates showed a smaller gap between more- and less-exposed work. In occupations with fewer graduates, the least-exposed jobs grew while the most-exposed declined.