The evidence cautions against claims of AI-driven graduate job collapse, but does not prove AI harmless either
The class of 2026 entered college around the time ChatGPT appeared. Now, as they enter the job market, graduates face two competing predictions. One says AI will eliminate the junior jobs that traditionally provide a first step into working life. The other says it will help inexperienced workers perform tasks that once required years of experience. What does the evidence show?
Our research examines US employment data through August 2026. We focus on people aged 22 to 25 whose highest qualification is a bachelor’s degree and who are no longer studying. Our main finding is that we do not detect a clear rise in unemployment among these graduates in summer 2026. But interpreting that finding requires care.
First, timing matters. Graduate unemployment typically rises in summer, when students finish their degrees and begin looking for work. Historically, summer unemployment rates have been almost two percentage points higher than during the rest of the year. A rise after graduation does not, by itself, show that AI is taking jobs.
Second, official unemployment figures miss some people who want work. Generally, someone without a job must have actively searched within the past four weeks to count as unemployed. Graduates who are moving, traveling or considering further study may not meet that requirement. Some may be struggling to enter employment; others may simply be taking a break. We therefore also examine a broader measure that includes people outside the labor force who say they want a job.
Comparing summer with summer, the official unemployment rate averaged 7.3% in 2026, against 7.2% in 2025 and 7.8% in 2024. Our broader measure reached 10.4%, the highest level across the five summers examined, compared with an average of 9.5% over the previous four.
That broader increase deserves attention. However, once we account for seasonal patterns, earlier trends and differences between individuals, neither measure shows an increase that can be reliably distinguished from statistical noise. We also compare young graduates with older graduates and young adults without bachelor’s degrees. These comparisons reveal no clear additional deterioration for young graduates.
Could national averages conceal problems in particular occupations? Among graduates with previous work experience, we find signs of higher unemployment in occupations more exposed to AI. However, these estimates are too uncertain to establish a clear relationship. These occupations also overlap substantially with jobs that can be done remotely, making competing explanations difficult to separate.
The evidence therefore supports caution about claims of an AI-driven collapse in graduate employment. It does not establish that AI is harmless. Weaker hiring in particular industries could remain hidden in national figures, and larger effects may emerge later. Unemployment alone also cannot tell us whether graduates are accepting worse jobs or lower pay.
For policymakers, universities and employers, the priority is to track these changes carefully. That means comparing equivalent seasons, following successive graduating classes and counting people who want work but fall outside official unemployment statistics. Graduates deserve advice grounded in evidence that distinguishes emerging problems from normal fluctuations.
© Robert W. Fairlie and Jane Wu
Robert W. Fairlie is Distinguished Professor of Public Policy and Economics at UCLA, US, and IZA@LISER Research Fellow
Jane Wu is Assistant Professor in the Strategy group at UCLA Anderson School of Management, US
Please note:
We recognize that World of Labour articles may prompt discussion and possibly controversy. Opinion pieces, such as the one above, capture ideas and debates concisely, and anchor them with real-world examples. Opinions stated here do not necessarily reflect those of the LISER.
Related World of Labour content:
https://wol.iza.org/articles/artificial-intelligence-and-labor-market-outcomes by Nick Drydakis
https://wol.iza.org/articles/innovation-and-employment-in-the-era-of-artificial-intelligence by Marco Vivarelli and Guillermo Arenas Díaz
https://wol.iza.org/articles/who-owns-the-robots-rules-the-world by Richard B. Freeman