AI could be the reason some young workers can’t find jobs

AI might be harming entry-level workers’ job prospects, at least in certain fields, new research suggests.

An as-yet unpublished study by three Stanford researchers found that, between late 2022 and July 2025, employment of people ages 22–25 “in the most AI-exposed occupations,” such as software developers, customer service representatives, and marketing and sales managers, fell by 6% overall. It dropped 13% compared with employment of people from the same age group in jobs with low exposure to AI.

But even in the most AI-exposed occupations, employment rose by 6%–9% for “older workers” (meaning, roughly, workers over 25) during the 2022–25 time period, the researchers found, suggesting that AI might be displacing certain entry-level jobs.

The researchers drew upon monthly payroll data for “tens of thousands” of companies from payroll provider ADP. To gauge how vulnerable a job might be to AI displacement, they looked at factors such as whether AI could perform tasks that fell under its description in the Labor Department’s Occupational Information Network (O*NET) database, as well as the number of Anthropic GenAI model Claude queries related to those tasks.

Their findings held true even when the researchers controlled for “firm-level shocks,” such as interest rate increases, that could have accounted for a drop in entry-level hiring. The pattern also held when the researchers excluded tech companies and “computer occupations” from the data, implying that AI, and not fluctuations in tech employment, was likely to blame for the decline in early-career employment.

So far, companies seem to be holding off on backfilling positions, rather than laying people off in favor of AI. It’s possible that entry-level hiring may pick back up if AI doesn’t perform as well as expected.

But if employers in certain sectors do find AI an adequate substitute for entry-level workers, the question remains: How will new entrants to those fields learn the skills they need to perform higher-level work?

Companies could find themselves with a bench-strength problem in years to come. “I think that we’ll have to more explicitly train people, as opposed to just hoping that they will figure these things out on their own,” Erik Brynjolfsson, one of the Stanford researchers, told the Wall Street Journal.