The rapid growth of generative AI, fueled by tools like ChatGPT since late 2022, has sparked global debate over its labor market impact — from optimism about productivity gains to fears of widespread job losses, especially for entry-level workers.
A recent paper, “Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence,” published by Stanford University provides evidence from millions of workers and tens of thousands of U.S. firms through July 2025, offering one of the clearest views yet of AI’s disruptive effects.
The study highlights key findings, with early-career professionals bearing the brunt of change:
Sharp Job Losses Among Young Workers in AI-Exposed Roles
Employment for workers aged 22–25 in AI-exposed occupations has fallen 13% since late 2022. Software developers and customer service representatives were hardest hit, with younger developers seeing nearly a 20% decline, while older peers in the same fields continued to grow.
Stagnant Job Growth Despite Strong Overall Employment
While the broader U.S. job market expanded, employment for 22–25-year-olds stagnated. In AI-exposed jobs, young workers saw a 6% drop, while mid-career workers (35–49) posted a 9% increase.
Declines Driven by Automation, Not Augmentation
The negative impact is concentrated in occupations where AI automates tasks entirely. In contrast, fields where AI augments human work — such as collaborative or knowledge-enhancing roles — have seen employment gains for young workers.
Persistent Effects Beyond Firm-Level Shocks
Controlling for firm or industry shocks, young workers in highly AI-exposed jobs still experienced a significant 12 log-point employment decline, confirming the trend is not explained by broader macroeconomic conditions.
Job Losses More Visible Than Wage Cuts
Compensation has remained largely stable across age groups and exposure levels. This suggests AI disruption is first reflected in employment opportunities, not wages, possibly due to wage stickiness in the short term.
Findings Robust Across Samples
The patterns hold even when excluding tech firms, remote-friendly roles, or outsourcing-sensitive jobs. The shift began in late 2022 — aligning with generative AI’s rise — and is evident across both high- and low-education occupations. Notably, even non-college workers up to age 40 in low-degree fields have faced employment divergence.
The evidence points to a disproportionate impact of generative AI on early-career workers, with job losses concentrated in roles most exposed to automation. For now, the disruption is evident in employment rather than wages, signaling an adjustment phase where displaced workers must transition into areas of new labor demand. As AI adoption deepens, continued monitoring will be critical to understanding its long-term effects on the workforce.
