Stanford Study: AI Widens Entry-Level Employment Gap to 19%

Summary

Stanford University economists’ August 2026 update to “Canaries in the Coal Mine?” found employment among workers aged 22 to 25 in the most AI-exposed occupations was 19 percent below less-exposed peers, widening from 13 percent last year. Since 2022, employment for this group fell about 11 percent in the top 40 percent of AI-impacted jobs while rising 10 percent in the least-impacted 60 percent. Economy-wide differences were minimal, and older workers remained largely unaffected. The divergence reflects weaker entry-level hiring, not increased firings, quitting or reduced pay.

Researchers analyzed anonymized, high-frequency ADP payroll data, scoring exposure through an existing labor-impact gauge and Anthropic’s Economic Index of workplace Claude use. Google published a similar occupational Gemini analysis the previous month. Automation-heavy occupations, including accountants, auditors, receptionists and information clerks, recorded the weakest entry-level employment, while augmentation-heavy roles such as chief executives and registered nurses showed flatter or rising employment. O*NET education requirements indicated that codified knowledge correlated with slower entry-level growth, while tacit knowledge favored mid-career and senior workers. Occupations with more college graduates showed smaller AI-related gaps. Lead researcher Erik Brynjolfsson said the effects are real, persistent and widening, risking a labor market that preserves existing jobs while removing career entry points.