Professional AI coverage has spent three years telling college students their degrees are about to be eaten by a model. The labor data, finally, says otherwise. In June 2026, graduates aged 22 to 27 sat at a 5.7 percent unemployment rate, according to the Federal Reserve Bank of New York. Yes, that is above the national average and a full point higher than two years ago. But an annoying 5.7 percent is not an apocalypse.
Summer graduate unemployment has wobbled between 6.3 and 7.8 percent since 2022, in other words since before ChatGPT even had a public preview. A fresh CESifo paper went hunting for "little direct evidence" that AI is shoving new graduates out of work, and by its own admission mostly failed to find it. Median monthly AI spending per employee has doubled since the end of 2025, yet there is still no statistically meaningful shift in who gets hired.
I do not buy the doom pile. But I also do not buy the smug counter that students are simply whining about nothing. The truth is more uncomfortable, and more interesting, than either headline.
The Real Crisis Is Boring, and Worse Than the Apocalypse
The dramatic version of this story is that AI is stealing entry-level jobs. The data says something quieter and harder to fix: AI is not replacing graduates so much as burying them behind a hiring machine that no longer looks at them as individuals.
Consider the underemployment numbers. Recent college graduates working in jobs that do not require a degree have reached 42 percent, against 33.7 percent for all graduates. That is the real story of the class of 2026: not idleness, but a brutal mismatch between a four-year credential and the work actually available.
Now layer AI on top. Two-thirds of recruiters said at the start of the year that they planned to expand AI pre-screening of interviews, per LinkedIn. Roughly eight in ten job seekers say they have used or plan to use AI in their own hunt. Meanwhile Revelio Labs, which tracks hiring, notes that the companies adopting AI fastest are also hiring fastest, except that they are prioritizing senior roles over junior ones. The firms are not cutting junior work because a model does it. They are quietly concentrating all the entry points on fewer, later-career hires. As Revelio's founder put it bluntly, 22-to-25-year-olds are getting lost in the shuffle.
That is the part the panic economies miss. Entry-level jobs were always a kind of subsidized apprenticeship, a way for young workers to buy experience they did not yet have. When companies tighten the aperture, the subsidy quietly disappears, and it has nothing to do with what a language model can generate. It has to do with risk, cost, and a screening system engineered to avoid people.
What I Would Tell a 2026 Graduate
I would start by refusing the two available scripts. You are not about to be automated away, and you are not helpless in a rigged game. The evidence lets you hold both truths.
Here is what the data says actually works:
- Stop submitting the same resume into a black box. Sixty-six percent of recruiters are screening with AI, and a resume built for a human reader is a resume that never gets read. Tailor every submission to the specific posting and its exact language.
- Show the work instead of claiming it. Handshake found the class of 2026 mentions AI skills on resumes at twice the rate of 2022, and the mentions that were tied to a real project were the ones that survived scrutiny. A portfolio beat a keyword list every time.
- Hunt the senior-heavy funnel's edge. If AI-forward firms are hiring, they are hiring for reason; the way in is often a demonstrable, applied skill rather than a generic degree line.
None of that is a cure-all, and I would not pretend it is. The NY Fed itself concludes that AI is not the main driver of the hiring slowdown, and its employer surveys show most companies intend to absorb AI through retraining, not layoffs. The ceiling on young careers right now is not the technology. It is an opaque gatekeeping system that has decided diversity of entry is expensive.
So here is the thesis I will actually defend: the college-grad AI apocalypse is a convenient fiction, but the gatekeeping that AI has accelerated is a real and fixable problem. You cannot fix it by panicking, and you cannot fix it by dismissing students as soft. You fix it the boring way, by demanding transparency about automated screening, feedback when a rejection happens, and a deliberate reopening of junior pathways at the companies doing the fastest hiring.
The models did not take your job, graduate. The door got harder to see. Those are not the same thing, and confusing them is the one mistake this job market cannot afford.
Comments