For years, AI watchers of all stripes have warned of a jobs apocalypse driven by systems able to replicate most human tasks more cheaply. Updated research from Stanford economists draws a narrower but sharper picture: AI appears to be causing significant entry-level job losses for younger workers in some fields, while older workers seem largely unaffected so far.

The August 2026 edition of "Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence" updates a paper of the same name with fresh data and refined statistics. The finding: employment levels for workers aged 22 to 25 in the most AI-exposed occupations are now 19 percent below those of peers in fields less exposed. A year ago that gap measured just 13 percent.

How it was measured

The researchers used a large subsample of the anonymized payroll data regularly aggregated by HR management company ADP. Each occupation's exposure to AI was rated on two separate measures:

  • A labor market impact gauge established by previous researchers — how much an occupation could theoretically be affected.
  • The Anthropic Economic Index — how occupations actually use the model in everyday work. Google released a similar report based on occupational Gemini usage in July 2026.

Invisible in aggregate, visible when separated

Crunching the numbers economy-wide, the researchers found little to no difference in relative employment between the jobs judged most and least affected by AI. The apocalypse in the headlines has no counterpart in the aggregate data.

Separate out workers aged 22 to 25, though, and the picture changes. Since 2022, employment in the top 40 percent of AI-impacted jobs has fallen by about 11 percent for that group. In the 60 percent of jobs with the least AI impact, employment for the same young workers grew by 10 percent over the same period.

Digging deeper reveals the mechanism: the loss comes not from increased firings or resignations but from lower hiring rates for entry-level workers. The effect shows up in employment volume rather than pay. Existing workers are not being shown the door; the door is simply opening less often for new arrivals.

Not all exposure is the same

The study's sharpest finding is that the word "exposure" hides two different things. Anthropic's index distinguishes uses that fully replace work previously done by a human from uses that help human workers be more effective at tasks they are still needed for.

By that measure, accountants and auditors, along with receptionists and information clerks, rank among the occupations most susceptible to automation. Chief executives and registered nurses, by contrast, are among those using AI most often in an augmentative way.

The result runs in the expected direction: jobs where automation is prevalent show the worst entry-level employment, while jobs where AI merely augments show a much more muddled picture. As the researchers write, the findings are "consistent with automation-oriented uses of AI substituting for labor while complementary uses are associated with flat or rising employment."

Why it matters

This research changes how the AI and employment debate should be conducted. The question is not whether AI will destroy jobs, but which end of which jobs it is destroying.

A narrowing entry level is a different problem from mass unemployment, and in some ways a more insidious one. Nobody loses a job, so no shock appears in the statistics. But the door into a profession narrows, and the consequence surfaces years later: the young person not hired today cannot be experienced ten years from now. The pipeline that feeds senior ranks dries up.