Stanford finds 14–20% drop in young hires
- Stanford economists Erik Brynjolfsson, Bharat Chandar, and Ruyu Chen say young workers in AI-exposed jobs saw a 16% relative employment drop after generative AI spread. (digitaleconomy.stanford.edu) - The pattern is concentrated among ages 22 to 25, and it shows up in employment more than pay — a hiring slowdown, not broad wage cuts. (digitaleconomy.stanford.edu) - The bigger point is simple: AI may be hitting the bottom rung first, even while older workers in the same fields stay stable. (digitaleconomy.stanford.edu)
The labor-market story here is not “AI is taking everyone’s job.” It’s narrower — and maybe more worrying. The first people getting squeezed look like young workers trying to get o(digitaleconomy.stanford.edu) 16% relative decline in employment, while older workers in those same occupations have stayed roughly stable. (digitaleconomy.stanford.e([digitaleconomy.stanford.edu)cts-of-artificial-intelligence/)) ### What is the actual finding? The core paper comes from Erik Brynjolfsson, Bharat Chandar, a(digitaleconomy.stanford.edu) how exposed they are to generative AI. Their headline result is a 16% relative employment decline for early-career workers in the most AI-exposed jobs after widespread GenAI adoption. The paper frames this as an early warning sign — “canaries in the coal mine” — rather than proof of an economy-wide collapse. (digitaleconomy.stanford.edu), the damage shows up there before it shows up in senior ranks. The Stanford paper says exactly that split appears in the data: young workers in exposed occupations fall back, while more experienced workers in those same occupations do not show the same decline. (digitaleconomy.stanford.edu) ### Is this about layoffs? Mostly, no. That’s one of the most important details. The Dallas Fed, discussing the Stanford result and checking i(digitaleconomy.stanford.edu)ddenly becoming unable to find anything at all. Basically, the front door is narrowing. (dallasfed.org) ### Why would AI hit juniors first? Junior jobs often bundle routine writing, research, coding support, customer communication, and document work — exactly the kind of digital tasks large language models can now do tolerably well. Older workers still carry tacit knowledg(digitaleconomy.stanford.edu)and use AI to reduce the need for the first rung beneath them. That mechanism is an inference, but it fits the age split the data shows. (digitaleconomy.stanford.edu) ### Could this just be high interest rates? The Stanford team pushed on that too. In a February 9, 2026 note, they argued that interest rates do affect o(dallasfed.org) drop in entry-level employment inside AI-exposed occupations. They also note the timing gets cleaner in 2024 once broader controls are added, which matters because it suggests some early weakness may have come from other forces mixed in with AI. (digitaleconomy.stanford.edu) ### So is the 14–20% claim right? Close, but the best-supported headline from the Stanford paper is 16%, not(digitaleconomy.stanford.edu)per itself is the 16% relative decline for 22-to-25-year-olds in the most AI-exposed occupations. If you want the clean version, use that. (digitaleconomy.stanford.edu) ### Why does this matter beyond one cohort? Because entry-level jobs are where people build skills, references, and the first line on the résumé. If AI sh(digitaleconomy.stanford.edu)ductivity overall and still leave the youngest workers with fewer ways to get started. (digitaleconomy.stanford.edu) ### Bottom line The cleanest read is not “AI destroyed the labor market.” It’s that AI may already be reshaping who gets hired first — and young workers appear to be taking the hit before everyone else. (digitaleconomy.stanford.edu)
