What the data says about new grads and the AI job market
If you graduated recently and it feels like your classmates are having wildly different experiences depending on what they studied, that’s not your imagination. The grad job market didn’t get uniformly worse. It got lumpy. Some fields barely flinched while others fell off a shelf, and the split lines up closely with how much AI can do the work.
The majors that got hit
The clearest picture comes from the field where you’d expect the pressure first. The Economist looked at how recent computer science and information science graduates are faring, and found their full-time employment dropped from about 70% to 55% over three years. That’s a big move for a group of degrees that spent the last decade as the safe, high-paying bet.
Compare that to a field like psychology, which the same analysis found stayed roughly flat over the same stretch. A psychology grad and a CS grad walked into very different markets, and the difference wasn’t ability or effort. It was exposure. When a chunk of the early-career tasks in a field can be handled by software, the door for beginners narrows, and computer science is close to the front of that line.
Stanford’s AI Index adds a sharper version of the same signal. It found software developers aged 22 to 25 down about 20% from 2024. Older developers in the same field didn’t see that drop, which tells you this is a young-worker problem more than a whole-occupation collapse. The people getting squeezed are the ones who haven’t built a track record yet, in exactly the field that used to hand new grads the most offers.
Why the technical fields feel it first
There’s a logic to why coding-adjacent degrees are where the strain shows up early. A lot of what a junior developer did in their first year was well-defined work: writing simple functions, fixing small bugs, converting a spec into code. That’s the kind of task current tools handle reasonably well, so it’s the first thing a manager can lean on software to cover instead of a new hire.
The messier parts of the job stick around. Figuring out what to build, catching the thing the model got confidently wrong, deciding a plan is bad before anyone writes a line: those still need a person. But those are also the parts a fresh graduate hasn’t practiced, because you used to learn them by grinding through the routine work first. Take away the routine rung and the ladder starts a step higher.
Lower-exposure fields don’t have this problem to the same degree. A lot of psychology-adjacent, care-adjacent, and people-heavy work resists being handed to a chatbot, so the entry point held up better. Field of study turned into a decent predictor of how rough the first job search would be, which is a genuinely new thing for the field. For most of the last decade the advice was the opposite, and the technical majors were the ones you picked precisely because the hiring was reliable.
None of this means a technical degree stopped being worth it. The people getting hired into those fields still tend to do well, and the tasks that survived are the higher-value ones. What changed is that the on-ramp got steeper, so the same degree that once came with a queue of recruiters now asks you to prove more before the first offer shows up.
The slowdown that predates AI
Before pinning all of this on the models, it’s worth being honest about the rest of the picture, because the grad slump has more than one cause. A meaningful share of what recent graduates are feeling is plain hiring weakness that would exist with or without a single chatbot.
PwC’s 2026 barometer put recent-graduate underemployment around 42.5% as of late 2025, meaning a large slice of grads who are working aren’t in jobs that use their degree. Some of that traces to AI reshaping which roles exist. A lot of it traces to the broader market, where hiring has been running well below its pre-pandemic pace for reasons that have more to do with interest rates and a post-2021 hangover than with anything a model can do. Both things are true at once, and it’s easy to over-credit the flashier explanation.
That matters for how you read your own situation. If the market is slow for structural reasons, that pressure eases as the economy loosens up, on a timeline that has nothing to do with your major. The AI-driven part of the shift is more permanent, and it’s concentrated in specific fields rather than smeared evenly across everyone.
How to read this if you’re the grad
I don’t want to soft-pedal it: if you studied something highly exposed and the search has been brutal, the data backs you up. But a few things are worth holding onto.
The tighter entry point is a near-term problem, not a verdict on the field. Demand for people who can actually work alongside these tools is climbing fast, and the developers who learn to direct the software rather than compete with it are in a good spot a couple of years out. The rung is higher, not gone.
What clears the higher rung is showing the judgment the routine work used to teach. Employers in these fields are increasingly reading for whether you can catch a mistake, reason through an ambiguous problem, and explain a decision, not just whether you can produce clean output. That’s a set of skills you can build and, importantly, demonstrate, which is a lot of what a strong interview is really testing for.
The honest read
The grad market fractured along field lines, and the fracture tracks AI exposure closely. Computer science and information science grads took the hardest hit, lower-exposure majors held steadier, and part of the whole thing is just a slow economy that will thaw on its own schedule. None of that is comforting if you’re the one sending out applications. What it does mean is that the disadvantage is specific and, in most cases, temporary, and the way through it is proving you can do the parts of the work the software still can’t.