Young workers are bearing the brunt of AI's impact on the labor market, according to newly updated research from Stanford University economists. Employment levels for workers aged 22 to 25 in the most AI-exposed occupations are now 19 percent below those of their peers in fields less exposed to AI disruption — a gap that measured just 13 percent when the researchers published their first analysis a year ago. The findings come from the August 2026 edition of "Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence," an update and revision of the team's influential 2025 paper. The findings were detailed in a report by Ars Technica. For anyone following the economics of the AI industry, it is among the clearest evidence yet that displacement effects are real — and concentrated.

The researchers, led by Stanford economist Erik Brynjolfsson and colleagues, found that the employment trends they identified last year are not only persisting but expanding.

The Numbers Behind the Gap

The study draws on a large subsample of anonymized, high-frequency payroll data aggregated by HR management company ADP. To rate each occupation's "exposure" to AI disruption, the team used two independent measures: a potential labor market impact gauge established by previous researchers, and the Anthropic Economic Index, which tracks how various occupations actually use Anthropic's Claude model in everyday work. Google released a similar report based on occupational Gemini usage last month.

When the researchers crunched the numbers economy-wide, they found little to no difference in relative overall employment between the jobs judged most and least affected by AI. The disruption only becomes visible when the data is broken down by age:

  • Since 2022, employment for workers aged 22 to 25 in the top 40 percent of AI-impacted jobs has fallen by roughly 11 percent.
  • Over the same period, employment for young workers in the 60 percent of jobs least impacted by AI grew by 10 percent.
  • The result is a large and widening divergence among early-career workers that simply does not appear among their older colleagues.

Hiring Freezes, Not Layoffs

One of the study's most important findings is how the gap is forming. According to the researchers, the effect manifests mainly through lower hiring rates for entry-level workers in AI-impacted fields — not through increased firings, layoffs, or people quitting. Young workers in exposed occupations are, in effect, not being replaced as their cohorts churn.

The researchers also found that the labor market effects among this age group showed up primarily as lower overall employment rather than reduced pay rates. People who have jobs in AI-exposed fields are not, on the whole, taking big pay cuts — there are simply fewer of them entering those fields.

That pattern is consistent with companies quietly re-routing entry-level work — tasks like drafting, summarizing, basic coding, and routine analysis — to AI systems, and shrinking the junior rungs of the career ladder as a result. It also matches what several major employers have said publicly about slowing graduate hiring in functions where AI tools now handle first-pass work.

Older Workers Largely Unaffected — So Far

Perhaps the most striking finding is what the study did not find: older workers appear largely unaffected so far. The employment gap is concentrated almost entirely among workers in their early twenties, the cohort that typically enters the workforce through exactly the kind of routine cognitive work that large language models now perform cheaply.

The asymmetry has uncomfortable implications. If AI absorbs the tasks that once trained novices, the traditional pathway by which junior employees become senior ones is weakened — even if mid-career professionals currently feel little direct pressure. Economists sometimes describe this as the "collapsed ladder" problem: the jobs disappearing are not careers, but the on-ramps to careers.

Caveats and Context

The findings come with important limitations. The effect sizes are relative, comparing AI-exposed occupations with less-exposed ones, not absolute unemployment. The exposure measures themselves are proxies — based on task surveys and model usage data — and cannot capture every nuance of how AI changes a given role. Some of the measured decline could reflect broader post-pandemic shifts in junior hiring, though the widening gap between exposed and unexposed occupations points specifically at AI as the differentiating factor.

The study also does not claim that AI reduces total employment across the economy. The aggregate effect measured economy-wide was muted; the damage is localized in specific occupations at specific career stages. New roles — AI operations, oversight, and integration work — are being created too, though not necessarily at the same pace or for the same people.

Still, the direction of travel is hard to ignore. A year ago, the same research showed a 13 percent gap and was debated as an early signal. Twelve months later, the gap has grown to 19 percent, driven by continued weak hiring for young workers in the most exposed fields. What began as a hypothesis about AI's labor market effects is hardening into a measurable trend — and the workers feeling it first are the ones with the least bargaining power.

---

Stay Ahead of AI

Get the latest AI news, analysis, and breakthroughs — all in one place.

Read more AI news →