AI hits newcomers hardest: youth employment in affected occupations down 19%

For years, watchers of the AI industry have predicted a “jobs apocalypse” to be wrought by superintelligent systems capable of doing most human tasks more cheaply. An updated study by economists at Stanford University shows that the blow is not falling on everyone at once, at least for now, but on the youngest.

The August 2026 revision of “Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence” refines last year's report with fresh data. The conclusion is the same, but the alarm has grown sharper: employment among workers aged 22–25 in the occupations most “exposed to AI” is now 19% lower than among their peers in AI-resistant fields. A year earlier, that gap stood at 13%.

The young versus the experienced

The authors drew on anonymized payroll data aggregated by the HR company ADP, and they gauged the “vulnerability” of occupations partly by the Anthropic Economic Index — an index of how various professions use the Claude model in their day-to-day work (Google put out a comparable report for Gemini).

Across the economy as a whole, the difference in employment between “vulnerable” and “resistant” occupations is almost imperceptible. But single out young people aged 22–25 and the picture changes: since 2022, employment in the 40% most affected occupations has fallen by roughly 11%, while in the remaining 60% it has risen by 10%. What's more, the decline is coming not through layoffs but through a drop in the hiring of newcomers, and it hits the number of positions rather than wages.

Replacement or assistance

The key distinction is between “automating” and “augmenting” uses of AI. Where the technology fully replaces a person (accountants, auditors, receptionists), youth employment falls the most. Where AI merely assists an experienced specialist (managers, nurses), the picture is calmer. “The results are consistent with automating uses of AI displacing labor, while augmenting uses are associated with employment growth,” the researchers write.

The authors also suggest that the most vulnerable occupations are those that rely on “codified” knowledge — formal, standardized knowledge that is taught from textbooks. By contrast, “tacit” knowledge, acquired through practice and mentorship, is for now only being augmented by AI. Higher education, meanwhile, serves as a buffer: in occupations with a large share of university graduates, the difference between vulnerable and resistant fields is smoothed over.

The study's lead author, Erik Brynjolfsson, states the concern plainly in an interview with The Washington Post: the jobs of those who entered a profession before the AI era will largely be preserved, but for the new generation many doors are beginning to close. “The effects on newcomers that we're measuring are real, persistent, and growing,” he says. “I'm increasingly worried about a labor market that holds overall employment steady but quietly shuts the door on those just starting out in their careers.”

Source: Ars Technica