Despite growing fears that AI will close the doors to entry-level jobs for university graduates, recent data indicates that this scenario has not clearly materialized so far.
اضافة اعلان
With the rising use of AI tools inside companies during 2026, new graduates were expected to be the first to be hurt, but a new study found that unemployment rates among them haven't recorded an exceptional jump so far, and haven't shown clear evidence of a broad decline in their hiring prospects.
The study is titled "Early Effects of AI on Employment Among New College Graduates," prepared by researchers Robert Ferley and Jane Wu, and published on September 15, 2026, as part of CESifo working papers. The study relied on US Current Population Survey (CPS) data, focusing on new bachelor's degree graduates aged 22 to 25 who are not pursuing graduate studies, in order to measure whether the accelerating use of AI has begun to actually be reflected in their unemployment.
Unemployment didn't jump in summer 2026
The study's most notable finding is that the unemployment rate among new graduates in summer 2026 reached 7.3%, a figure that doesn't fall outside the range recorded in previous years. The rate was 7.1% in 2022, 6.3% in 2023, 7.8% in 2024, and 7.2% in 2025.
The data also showed that unemployment rose seasonally during the summer months, a usual pattern with new cohorts entering the labor market. In 2026, the rate rose from 5.4% in May to 7.8% in June, then fell to 7.4% in July and 6.8% in August. Based on this, the study sees the 2026 situation as not looking anomalous compared to the summers of previous years.
No evidence of a broad decline in hiring
The study stresses that the findings don't point to a major exodus or broad decline in new graduate hiring, whether measuring the situation in absolute terms or comparing it to other groups.
For this purpose, the researchers compared new graduates' situation with two groups: young people of the same age but without a college degree, and older college graduates, aged 30 to 49. In most of these comparisons, the study didn't find statistically significant differences indicating that 2026 graduates experienced a clearly different unemployment shock due to AI.
Broadening the definition of unemployment didn't change the picture
The study didn't settle for the official definition of unemployment, but added another category of young people who said they wanted to work but weren't actively looking, and thus aren't officially counted among the unemployed.
This broadening raised the unemployment rate among new graduates by around two percentage points, but didn't change the core finding: there was no statistically significant increase in summer 2026 even after including this category, which the study described as "marginalized unemployed" or those "excluded from official measurement."
Why were new graduates the study's focus?
The researchers chose to focus on new graduates because they're the group most exposed to any early shift in labor demand. Companies, if they want to benefit from AI, may begin first by reducing new hiring for simpler, more standardized jobs, rather than laying off experienced employees.
This makes new graduates an early indicator of any actual change in the labor market. Summer 2026 was also seen as the first hiring season in which broader effects of AI use in workplaces might appear.
Why does the result differ from the Stanford study?
This study comes with a result that appears different from an earlier Stanford University study that spoke of a clear weakness in hiring prospects for beginners in the jobs most exposed to AI.
This difference stems from the nature of the data used. The Stanford study relied on payroll and employment data from ADP, and focused on the volume of hiring within different occupations. The CESifo study, meanwhile, relied on unemployment data from the Current Population Survey, meaning it measures graduates' situation in the labor market from a different angle, also encompassing overall demand for jobs and seasonal factors.
In other words, signs of weakness may appear in certain jobs or occupations, without yet being reflected as an unusual rise in unemployment among new graduates across the market as a whole.
Some signals appeared... but without a conclusive answer
The study also analyzed whether unemployment in 2026 was rising more in jobs with higher exposure to AI, or in jobs more amenable to remote work.
The results showed that the relationship with AI exposure indicators was positive but not statistically significant, while a limited positive relationship appeared with jobs more amenable to remote work. Therefore, the researchers concluded that the data doesn't yet allow isolating a clear, direct effect of AI from the other changes affecting new graduates' circumstances.
The study's core takeaway appears to be that AI hasn't yet shown a broad, clear effect on unemployment among new US college graduates, at least according to summer 2026 data.
But the researchers don't consider this a final verdict, only a preliminary test. They warn that continued expansion of AI use in workplaces could make the 2027 cohorts and those after more exposed, meaning the final judgment requires additional data over the coming years.
Resource: Al-Ghad.
