If you graduated this year and the job hunt feels harder than it should be, you are not imagining it. New data out of Stanford puts real numbers behind something young job seekers have felt since 2023: the entry level ladder in AI heavy fields is getting shorter, and it just got shorter again.

The AI job gap for new graduates is not a vague headline this time. It comes from a team at Stanford’s Digital Economy Lab that tracks real payroll data every month, and their latest update shows that gap widening in a way that is hard to argue with.

The AI Job Gap for New Graduates: What the New Data Shows

Economists Erik Brynjolfsson, Bharat Chandar, and Ruyu Chen run an ongoing study called “Canaries in the Coal Mine”, built on ADP payroll records covering millions of US workers. Their August 2026 update, tracked live on the public Canaries Dashboard, found that workers aged 22 to 25 in the most AI exposed occupations are now about 19 percent below where their employment would be if it had kept pace with less exposed peers the same age. A year earlier, in July 2025, that gap was 15 percent. It is not closing. It is opening wider.

In plain numbers, employment for 22 to 25 year olds in the two most AI exposed job categories fell about 11 percent between November 2022 and June 2026. For the same age group in the three least exposed categories, employment grew around 10 percent over that same stretch. Experienced workers in the exposed roles show no comparable drop at all.

Why This Is Hitting New Grads, Not Everyone

The researchers are careful to say this is not mass layoffs. Companies are not firing junior staff en masse. The adjustment shows up almost entirely in hiring: businesses are opening fewer entry level seats in roles where AI tools can now do a chunk of what a junior employee used to do, things like first drafts, basic research, routine code, and simple analysis.

The split that matters more than the headline number is this:

• Jobs leaning on codified knowledge: formal, documented, teachable skills found in manuals and training material. AI tools are strong here, and these roles are losing entry level ground.
• Jobs leaning on tacit knowledge: judgment built from experience, reading a room, handling an unusual situation. These roles are flat or growing, for junior and experienced workers alike.

So the real question is not “will AI take my job.” It is closer to “does my job mostly ask me to follow a known process, or does it ask me to use judgment that is hard to write down.” The second kind of work is holding up far better.

From my own experience running websites and online tools, I have watched this play out in miniature. The tasks I used to hand to a junior freelancer, formatting, basic edits, first pass research, are the ones AI now handles for me directly. The tasks that still need a real person are the ones where someone has to make a judgment call I cannot fully explain in a brief.

Quick tip: when weighing two entry level roles, ask how much of the job is “follow the process” versus “use your judgment on something messy.” The messier, more judgment heavy role is more likely to still need a person in five years.

What This Means for You Right Now

This does not mean entry level jobs are disappearing everywhere. It means the templated end of junior work is shrinking faster than the judgment heavy end. A few moves that track with what the data actually shows:

• Target roles where the job description leans on client handling, troubleshooting, or decisions under uncertainty, not just task execution.
• Learn to work with AI tools directly instead of competing with them on tasks they already do well. Being the person who checks, directs, and fixes AI output is more durable than being the person who does the raw task by hand.
• Build a visible portfolio of judgment calls you made, not just outputs you produced.
• Look at fields the data shows holding up: cybersecurity, hands on healthcare support roles, and skilled trades all lean heavily on tacit, situational knowledge.

We covered the bigger picture in Is AI Taking Entry-Level Jobs? What New Graduates Should Know, and this new Stanford update confirms the trend described there has kept moving, not reversed. For the wider decade long view, our piece on what McKinsey’s new report means for your career pairs well with this one.

If you want to stand out while this shakes out, something you can actually show matters more than another resume line. Our guide on how to build an AI portfolio for free walks through that, and AI skills on your CV: what the research says they are actually worth is a useful check before you list anything that sounds impressive but means little to a hiring manager.

Common Questions

Is this just one study, or is it widely accepted?
It is one of the most carefully built studies on this question, using real payroll data rather than surveys, updated monthly through a public dashboard. The researchers call it an early warning system, not a final verdict.

Does this mean experienced workers are safe?
The data shows no comparable gap for experienced workers in the same roles, but that reflects hiring patterns today, not a guarantee for the future. Skills still matter more than years on the job.

What industries show the widest gap?
The study measures exposure using a methodology from Eloundou, Manning, Mishkin, and Rock’s research on LLMs and labor market exposure. The pattern holds across jobs heavy on codified, documented tasks, think routine coding, basic research, and first draft writing, rather than any one named industry.

Should I avoid learning AI tools since they are taking entry level jobs?
No, the opposite. Jobs where AI complements a worker’s judgment are growing, not shrinking. Learning to direct and check AI tools is becoming a baseline skill rather than an optional extra.

The Bottom Line

The AI job gap for new graduates is real, it is measured in actual payroll data, and it widened again this year. But it is a hiring slowdown in specific kinds of roles, not a collapse of the entry level job market. The graduates doing best right now are aiming at judgment heavy work and treating AI tools as something to direct, not compete against. That is more useful than either panic or denial, and it is something you can act on this week.