Stanford's youth-employment gap in AI-exposed jobs widened to 19%

Stanford's Digital Economy Lab published a revised version of its ongoing employment analysis on August 12, 2026, led by economist Erik Brynjolfsson with Bharat Chandar and Ruyu Chen, using ADP payroll records covering roughly 4.6 million workers from November 2022 through June 2026. Its central finding: employment among workers aged 22 to 25 in highly AI-exposed occupations now sits about 19% below where it would be had it kept pace with employment among similarly aged workers in less-exposed occupations -- up from a 15% gap the same team measured a year earlier, in the July 2025 vintage of this data.

The researchers describe the mechanism as running through hiring, not firing: young workers in exposed fields aren't being let go at a higher rate, they're being hired into those roles at a lower one, and the effect is concentrated in occupations -- software engineering and customer service among them -- that lean on "codified knowledge," the kind that can be learned from documentation and training text rather than judgment built on the job. Older, more experienced workers in the same occupations show no comparable gap.

A narrower claim than the headline version

The Lab's own text is explicit on the scope of the claim, in a line quoted directly in this update: "We do not see widespread, economy-wide job displacement associated with AI." What they do see is this specific, narrower pattern -- a widening gap concentrated in young workers and in a particular category of occupation -- and the report presents it as exactly that, not as evidence of a broader labor-market collapse. For a site built around measuring what AI spend actually produces rather than what people assume it's doing, that's a useful discipline to borrow: a real, worsening trend among one group of workers, reported without inflating it into a claim the data doesn't support.