A Stanford Digital Economy Lab study published on 12 August 2026 found that employment among workers aged 22–25 in AI-exposed occupations in the United States is about 19% below where it would be had it tracked less-exposed peers, with the gap widening steadily since researchers first documented it in August 2025.
The revised paper, titled ‘Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence,’ was authored by Erik Brynjolfsson, Bharat Chandar, and Ruyu Chen. It draws on high-frequency administrative payroll data from ADP covering millions of U.S. workers through June 2026, using November 2022 as the baseline before the widespread rise of generative AI in the labour market.
The researchers stressed that they found no evidence of widespread, economy-wide job displacement from generative AI. However, they identified a sharp and growing divergence for young workers in roles where AI is more likely to automate rather than augment human labour.
However, employment of young workers (ages 22–25) in AI-exposed occupations now stands 19% below where it would be had it kept pace with that of their less-exposed peers,
Stanford Digital Economy Lab authors Brynjolfsson, Chandar, and Chen wrote in their 12 August 2026 summary. They noted that the divergence has widened steadily since the first documentation in August 2025, when Axios reported a 16% decline for the same age cohort, and an earlier Stanford working paper reported a 13% relative decline after controlling for firm-level shocks.
The study says the employment shortfall among young workers is driven mainly by reduced hiring rather than increased separations or layoffs. This means companies are simply not bringing in new entry-level talent at the same rate in AI-exposed occupations, even as older workers in those same roles have not shown a comparable decline.
Our findings are consistent with the hypothesis that AI is having this effect on the labor market, especially for entry-level workers.
Bharat Chandar told Time magazine. The researchers describe the results as early, descriptive indicators rather than causal estimates, meaning the study establishes a strong correlation but has not definitively proven that AI alone is responsible for the hiring decline. Ars Technica, reporting on the findings on 24 August 2026, identified accountants, auditors, receptionists, and information clerks as among the occupations most susceptible to AI automation, while chief executives and registered nurses were among jobs using AI augmentation most often.
The findings carry potential implications for South African graduates entering the labour market, where youth unemployment among those aged 15–34 has long been among the highest in the world according to Statistics South Africa’s Quarterly Labour Force Survey. If AI-driven automation begins suppressing entry-level hiring locally in sectors such as accounting, administrative support, and customer service, the impact on young job-seekers in Cape Town and across the country could be significant. However, no South Africa-specific labour-market effect has been verified by the Stanford study or any supplied source, and the findings are drawn entirely from U.S. payroll data and selected AI-exposed occupations.
This is the fastest, broadest change that I’ve seen,
Erik Brynjolfsson told Axios in August 2025, comparing the pace of AI-driven labour market shifts to previous technological transitions. The figures have shifted across versions of the analysis — from 13% in the early working paper to 16% reported by Axios in August 2025 to the current 19% in the revised August 2026 paper — reflecting updated data vintages rather than contradictory findings.
What remains unknown is whether the trend will extend beyond the United States to labour markets in developing economies such as South Africa, where the structure of entry-level work and the pace of AI adoption differ substantially. CapeFlats.co.za will check for any South African labour-market data or local studies examining AI’s impact on graduate hiring when next updating this story in October 2026.