September 20, 2026
AI-driven layoffs have not materialized: Yale found no effect 45 months after ChatGPT
AI's effect on employment and wages in the US is indistinguishable from zero: the Yale Budget Lab measure now includes data through August 2026.
US unemployment rose from 3,4% in April 2023 to 4,3% in March 2026. AI is not visible in these figures.
The Budget Lab at Yale has tracked the US labor market since November 2022, when ChatGPT launched. The lab found no mass worker displacement: AI's effect on employment and hourly wages in occupations where it can take on the most tasks is indistinguishable from zero. Net job growth has slipped to 20 000 a month, but the lab's calculations say this slowdown is not about AI either, so far.
Compared it with the internet. From 1996 to 2002, the occupational mix of US employment diverged by 7 percentage points: that is how many workers would have to move to different occupations for one profile to match the other. After ChatGPT launched, the trajectory runs about 1 point higher, and the divergence began as early as 2021, before generative AI.
Others found an exception. Stanford Digital Economy Lab, using ADP payroll data, estimates that employment among 22–25-year-olds in the occupations most exposed to AI is 19% below the trajectory of their less-exposed peers (revision from 12.08.2026; in the July 2025 snapshot, the gap was 15%). There is no gap among older workers, and the adjustment comes through reduced hiring of young people rather than layoffs.
The count understates programmers. The Budget Lab authors call this limitation out directly: developers were placed in the same occupational group as office clerks, but tools were rolled out quickly and at scale for developers, while clerical work lags behind. Averaging across the group hides an effect where it may have happened first.
Over 11 months of 2025, US companies attributed 55 000 announced job cuts to artificial intelligence, 4,5% of all cuts. They attributed 245 000 to market and economic conditions (Challenger, Gray & Christmas via Oxford Economics).
Marta Gimbel of Budget Lab explains the gap this way: a company will not say, “we got our economic forecast wrong”; a company will say, “the world is changing quickly, we are optimizing headcount.”
Budget Lab recalculates its tracker with each monthly employment report, and the latest one, from 15.09.2026, did not move the line.
Source
