Are new college graduates losing jobs to AI?

A new working paper from researchers at CESifo examined employment data from the U.S. Census Bureau to see if those fears were playing out in the real world. By tracking the unemployment rates of people aged 22 to 25 with bachelor degrees, they looked for signs that these graduates were being pushed out of the job market. Despite the high-profile predictions from business leaders and tech investors that 2026 would see a spike in youth unemployment due to AI, the actual data tells a different story. The unemployment rate for new graduates this past summer was 7.3 percent. This number is not an outlier; it sits comfortably within the range of 6.3 percent to 7.8 percent observed over the last four years. Across every group the researchers compared, there was no sign of a significant, widespread drop in hiring.

It is helpful to think of a company as a machine with different gears. Some gears, like senior managers, handle complex, high-level strategy. Others are entry-level roles, which often involve collecting information, checking facts, or formatting reports. AI is currently very good at these entry-level tasks because they are usually standardized—meaning they have clear, repeatable rules. The theory was that if you have an AI tool that can do the work of three juniors, you would hire zero juniors. However, labor markets are complex. Even if an AI can write a report, it still needs a human to verify it, context-check it, and communicate the results to the rest of the team. We may not be seeing mass displacement yet because companies are still learning how to integrate these tools, or because the demand for human employees is still high enough that firms are using AI to make their current staff faster rather than replacing them entirely.

This data serves as a reality check for the narrative that AI will replace entry-level workers overnight. While it is true that many companies are spending significantly more on AI software and token usage—the units of text or data that AI systems process to function—this has not yet translated into a measurable hiring crisis for young people. It is a reminder that there is a difference between a technology being capable of a task and a business being ready to overhaul its entire workforce. We are in the early days of this transition, and it is possible that the real impact will take years to show up in national statistics. For now, the most important takeaway is that while the tools of work are changing rapidly, the fundamental need for human employees remains steady.