A new analysis of US labor market data challenges the increasingly popular narrative that artificial intelligence is decimating opportunities for the newest entrants to the American workforce. While executives and investors have warned that generative AI is finally capable of replacing entry-level employees, a working paper from Munich’s CESifo research institute found that unemployment among recent college graduates in the summer of 2026 was not unusually high compared with the previous four years.
The study, titled “The Early Impacts of AI on Employment Among Recent College Graduates,” examined bachelor’s degree holders aged 22 to 25 who were not pursuing further education. Researchers Robert Fairlie and Jane Wu found that this group’s unemployment rate stood at 7.3 percent during the summer of 2026. That figure sits within the range observed since 2022, when the rate was 6.3 percent, and actually comes in below the 7.8 percent recorded in 2024.
The findings run counter to a recent study from Stanford University, which concluded that employment trends for younger workers in occupations considered heavily exposed to AI were weakening. That divergence stems partly from the different data sources used by the two research teams.
Fairlie and Wu focused specifically on new graduates because hiring reductions may be among the first ways that AI affects employment. Companies can limit new openings for routine office work without laying off experienced employees already on staff. Many entry-level roles involve tasks that are increasingly within reach of generative AI systems: summarizing documents, preparing first drafts, producing basic analysis, processing information and handling other standardized work. As these systems improve, employers may be able to assign part of that work to software rather than new hires.
The researchers drew on microdata from the Census Bureau’s Current Population Survey, which tracks employment and unemployment across the US workforce. They examined summer results from 2022 through 2026, a period that began after the labor market had largely recovered from the pandemic and included the release and rapid adoption of ChatGPT.
To test the robustness of their findings, the authors broadened the analysis to include people who said they wanted jobs but were not actively looking. That adjustment did not materially change the result. They then compared recent graduates with workers of the same age who did not have college degrees, as well as older college graduates aged between 30 and 49. The paper also categorized jobs by their estimated exposure to AI, using earlier research on the occupations where AI systems could potentially handle more tasks.
Those comparisons did not show statistically significant differences in employment trends during the period studied. “Unemployment among recent college graduates in summer 2026 was not unusually high relative to earlier summers,” the authors wrote. The data “tell a consistent story” across comparison groups, they added.
The finding does not mean that AI has had no effect on the job market. It means the effect is not yet clear in this particular measure of employment.
The contrast with the Stanford research may partly reflect the data each team used. Stanford relied on payroll records from HR company ADP to measure changes in employment across occupations. That can show whether employers are adding or cutting jobs in specific fields. The CESifo paper measures unemployment instead, capturing both job availability and the number of people seeking work. A company could reduce entry-level hiring in an AI-exposed field, for instance, while unemployment remains stable if workers find jobs elsewhere or fewer people seek work in that area.
Still, there are reasons to watch the next several graduating classes closely. The CESifo researchers noted that more companies have reported using AI to automate employee tasks. Spending on AI per worker has increased, and use of ChatGPT Enterprise has also grown over the past year.
Prominent voices in finance and technology have begun to issue similar warnings. Venture capitalist Marc Andreessen said earlier this year that “AI literally until December 2025 was not actually good enough to do any of the jobs that they’re actually cutting.” BlackRock CEO Larry Fink has also said that “when this year’s college graduates enter the workforce, we could see the highest unemployment rate among them in years – even without a recession.”
So far, broad Census data do not support that outcome for the class of 2026. But the CESifo researchers said it may be too soon to draw firm conclusions. If companies keep expanding AI use, graduates entering the labor market in 2027 and beyond could face a different hiring environment. The summer 2026 data serve as a “useful first test” of how accelerating AI usage is impacting the current US job market, the authors concluded, adding that “additional years of data will be needed to hone in on whether effects emerge as workplace use of AI deepens.”