The study records lower employment among workers aged 22-25 in sectors most exposed to AI
Stanford University updated its study “Canaries in the Coal Mine?” in August 2026, identifying an employment gap between workers aged 22-25 in occupations most exposed to AI and their peers in less exposed sectors.
For the analysis, researchers used a subset of anonymized, aggregated payroll data from ADP. They classified occupations using a measure of potential impact on work and the Anthropic Economic Index, which considers the use of Claude in professional tasks.
According to the authors, the pattern mainly concerns fewer hires for entry-level positions, rather than layoffs or resignations. Among young workers, the effects appear more in overall employment than in wages. Across the economy, differences between occupations classified by AI exposure are more limited.
The research distinguishes between automating use, which replaces work performed by people, and augmenting use, which assists workers in tasks they continue to perform. Occupations where automation prevails show relatively weaker employment outcomes for people at the start of their careers. Accountants, auditors, receptionists and information clerks are among the occupations most susceptible to automation.
The authors link the finding to codified knowledge, formalized in procedures, texts and education. Using the education requirements in the O*NET database, the study associates occupations with more codified knowledge with slower growth in entry-level employment. Occupations with more practical knowledge acquired through experience and mentoring instead show faster growth for mid- and late-career workers.
