Stanford Study Finds Experience May Shield Workers From AI Displacement

A new update from Stanford University suggests that artificial intelligence is creating a growing employment challenge for younger workers, even as there is still no clear evidence of widespread, economy-wide job displacement. The research found that workers between ages 22 and 25 in occupations highly exposed to AI are increasingly falling behind peers in less-exposed fields, while more experienced employees have not experienced a similar decline.

The employment gap for younger workers in highly AI-exposed occupations has expanded from 15% to 19% over the past year, according to the researchers. The change appears to be driven primarily by weaker hiring of younger employees rather than by an increase in workers losing their existing jobs.

The findings update a study published a year earlier under the title “Canaries in the Coal Mine?” The research was co-authored by Erik Brynjolfsson, Bharat Chandar and Ruyu Chen and draws on payroll information from ADP along with employment data from the U.S. Bureau of Labor Statistics.

When the researchers examined employment across occupations and age groups, they found no broad decline that could be characterized as an economy-wide wave of AI-related job losses. Instead, the effects were concentrated among younger workers in occupations with greater exposure to artificial intelligence. Experienced workers, meanwhile, showed no comparable employment deterioration, with some groups experiencing flat or increasing employment.

The study also identifies an important difference between codified knowledge and tacit knowledge. Younger workers have experienced employment declines in occupations that depend heavily on codified knowledge, meaning information that is formalized, standardized and documented and can typically be learned through education, written procedures or textbooks.

The pattern is different in occupations where workers rely more heavily on tacit knowledge developed through practice, mentoring and repeated exposure to real-world situations. Employment among experienced workers in these areas has increased, according to the research.

The researchers say this distinction may help explain why experience appears to matter in the current labor market. Generative AI is particularly capable of reproducing and applying information that has already been captured in text and other digital formats, while knowledge gained through direct experience remains more difficult to replicate.

The researchers caution, however, that the data cannot establish that generative AI is solely responsible for the employment differences. Other forces are also affecting the labor market, and the study cannot determine exactly how much of the divergence is attributable to AI. The researchers accounted for factors including changes in interest rates, education levels and remote-work trends before assessing the employment patterns.

The findings nevertheless suggest that AI may be contributing to shifts in employment among certain occupations. The researchers describe the evidence as consistent with a possible mild slowdown in employment growth for jobs most exposed to AI rather than a broad collapse in employment.

Payroll data from ADP showed average employment across the sample increased by about 6% between November 2022 and June 2026. Employment in the most AI-exposed fifth of occupations grew by approximately 4% over the same period. While that represents slower growth, it does not show widespread employment declines among AI-exposed occupations.

The researchers therefore found no evidence in the payroll data through June 2026 of an economy-wide “job apocalypse” caused by artificial intelligence. Stanford plans to continue tracking the labor-market effects of AI through its new AI Economic Indicators, providing additional data as employment patterns evolve.