Short answer: Not across the board, at least not yet. A new study of summer 2026 found no broad jump in unemployment for recent college graduates. But two other 2026 studies found that graduates from the majors most exposed to AI, like computer science, are finding first jobs more slowly and earning less than their peers.
Key takeaways
- Unemployment for recent grads aged 22 to 25 averaged 7.3% in summer 2026, inside the range of the past four summers, according to economists Robert Fairlie and Jane Wu.
- A Dallas Fed study of Texas graduates found that more AI-exposed majors saw lower job-finding rates and about 5% lower first-year earnings after ChatGPT arrived.
- A Census Bureau working paper found grads of the most exposed majors earned 13% less in their first three months and were shifting into lower-paying sectors like retail and food service.
- The studies do not really conflict. AI looks like a squeeze on certain majors, not a wave of unemployment for all new grads.
Why are people worried about AI and new college grads?
First jobs often involve drafting, summarizing, basic coding and data cleanup, tasks AI tools can now partly handle. If employers need fewer beginners, the class that just graduated would feel it first.
That worry has been loud. BlackRock CEO Larry Fink, quoted in the UCLA Anderson Review, warned that “we could see the highest unemployment rate among them in years.” In recent weeks, three research teams released evidence on whether that is happening. Here is what each one found.
Did unemployment for new grads spike in 2026?
What happened: In September, UCLA researchers Robert Fairlie and Jane Wu published a working paper through the National Bureau of Economic Research titled “The Early Impacts of AI on Employment among Recent College Graduates.” They used the federal Current Population Survey, the same survey behind the monthly jobs report, to look at college graduates aged 22 to 25 who are no longer in school.
Their main number: unemployment for this group averaged 7.3% from June through August 2026. That was higher than in 2022 (7.1%), 2023 (6.3%) and 2025 (7.2%), but lower than 2024 (7.8%). They also counted “sidelined” grads, people who want a job but stopped looking. That broader rate was 10.4%, the highest of the five summers, but the rise was not statistically significant. Unemployment in more AI-exposed jobs was slightly higher, but again the gap was small and not statistically significant.
Why it matters: This is the first look at the class of 2026, and it does not show the jobs collapse some predicted. As the UCLA Anderson Review summarized, summer 2026 “offers little evidence of a broad AI-driven unemployment increase.”
The other side: The authors call their results descriptive, not proof of cause. Older workers and young non-graduates, their comparison groups, could be affected by AI too. And they note that larger effects could still show up in later graduating classes.
Are AI-exposed majors like computer science doing worse?
What happened: Two studies say yes. On September 22, Dallas Fed economists Samuel Dodini and Tucker Smith published an analysis of graduates of Texas public four-year universities, linked to state wage records. They measured each major’s AI exposure using job postings from before ChatGPT launched, combined with a task-based measure built by Anthropic, the company that makes Claude.
After ChatGPT’s release, graduates from more exposed majors became less likely to find work in Texas within a year, compared with less exposed majors. The authors report a 1.7 percentage point relative decline in employment rates for more exposed majors. First-year earnings for more exposed majors fell by about 5% from 2021 to 2024, relative to other majors. The most exposed majors include computer science, computer engineering and languages. Nursing, education and psychology are among the least exposed.
The second study is a U.S. Census Bureau working paper, “Graduating into Disruption: Labor Market Outcomes for AI-Exposed College Majors.” As reported by ThePrint, it used records covering about 29% of U.S. bachelor’s degrees awarded from 2016 to 2024. Graduates of the 10% most AI-exposed majors were about 5% less likely to be employed within three months of graduating, and earned 13% less in those first months. Roughly half of the pay gap came from lower pay within industries. The rest came from grads moving into lower-wage sectors. ThePrint quoted researcher Lee C. Tucker saying these graduates “are shifting into retail, trade, and food and accommodation sectors.”
Why it matters: These studies follow real records, not survey answers, and two different data sets point the same way: AI exposure is linked to a rougher start for certain majors.
The other side: The Census researchers stressed that the study “does not suggest that highly exposed degrees or professions will disappear,” and the gap narrowed as graduates moved further into their careers. The Dallas Fed data only covers jobs in Texas, and its findings describe associations, not proven cause. Neither study covers the class of 2026.
How do the three studies compare?
| Study | Who it covers | Main finding |
|---|---|---|
| Fairlie and Wu (NBER, Sept. 2026) | U.S. college grads aged 22 to 25, summer 2026, survey data | Unemployment of 7.3%, within the range of 2022 to 2025. No broad AI spike. |
| Dodini and Smith (Dallas Fed, Sept. 22, 2026) | Texas public university grads, 2021 to 2024 classes, wage records | More exposed majors: lower job-finding rates and about 5% lower first-year pay. |
| Census Bureau working paper (Sept. 2026) | About 29% of U.S. bachelor’s degrees, 2016 to 2024 | Top 10% most exposed majors: 13% lower pay in first three months, shift to lower-wage sectors. |
So which is it? Is AI hurting new grads or not?
Both findings can be true at once. An average can look normal while one neighborhood gets hit by a storm.
The NBER study looks at all recent grads together. Grads of the most exposed majors are only one part of that group, so trouble for them can be hidden inside a normal-looking average. The Dallas Fed and Census studies zoom in on major-by-major differences, which is exactly where AI would show up first. They also show that the damage often looks like a worse job, not no job. A graduate working at a store instead of a tech firm counts as employed in the unemployment rate.
This fits our look at HubSpot’s layoffs: the clearest AI effect so far is fewer doors opening for young workers, not mass firings.
The TrustNoRobot Risk Check
| Question | What we see |
|---|---|
| Exposure | Highest for new grads in computer science, computer engineering, math, statistics, languages and journalism. Lowest in nursing, education and psychology. |
| Evidence | Moderate and growing. Two large record-based studies agree on a major-level effect. A survey study finds no broad unemployment spike. None prove cause. |
| Speed | Gradual. Effects appear in classes graduating since ChatGPT’s late 2022 release, and the class of 2026 has not yet been studied with wage records. |
| Cushion | Partial. The Census study found pay gaps narrow over time. But the Dallas Fed found master’s degree holders in exposed fields saw similar pay declines, so more school is not a sure fix. |
The other side
It is hard to separate AI from everything else going on. AI-exposed jobs are often the same ones that can be done remotely. Fairlie and Wu found unemployment was somewhat higher in remote-capable jobs, and they note that remote work and AI exposure overlap so much that the two are hard to pull apart.
There is also a measurement question. “Exposure” means a job’s tasks overlap with what AI can do. It is not a forecast of job losses. And a 5% or 13% pay gap, while real, is a setback, not a closed door. Most graduates in these fields still find work.
What this means for you
If you are a student or recent grad in an exposed major, this is a reason to plan, not panic. Build skills in the parts of your target job AI does poorly: judgment, client work, debugging real systems, and combining your field with another. Internships and real projects matter more when employers hire fewer beginners.
If you are a parent or mentor, do not steer someone away from a field based on one headline. The studies show a rougher start, not a dead end. Look up specific occupations in our Job Exposure Explorer, and walk through our five-step guide to measuring AI risk to your job together.
Frequently asked questions
Is AI causing unemployment for recent college graduates?
Not broadly, so far. A 2026 NBER study found summer unemployment for grads aged 22 to 25 was 7.3%, within the range of the four prior summers. Effects show up mainly in specific AI-exposed majors.
Which college majors are most exposed to AI?
Studies from the Dallas Fed and the Census Bureau list computer science, computer engineering, applied math, statistics, languages and journalism among the most exposed. Nursing, education and psychology are among the least exposed.
Do AI-exposed graduates earn less?
Yes, at the start. The Dallas Fed found about 5% lower first-year earnings for more exposed majors in Texas. A Census Bureau paper found 13% lower pay in the first three months for the most exposed majors, with the gap narrowing over time.
Will grad school protect me from AI?
Not reliably. The Dallas Fed found master’s degree holders in more exposed fields saw pay declines similar to bachelor’s graduates. Skills that complement AI may matter more than extra credentials.
Sources
- Fairlie and Wu, “The Early Impacts of AI on Employment among Recent College Graduates,” NBER Working Paper 35796 (Sept. 2026)
- Fairlie and Wu, IZA Discussion Paper No. 18945 (full text) (Sept. 2026)
- UCLA Anderson Review, “For New College Grads, the AI Jobpocalypse Didn’t Arrive” (Oct. 7, 2026)
- Dodini and Smith, “AI plays a role in weak labor market for college graduates,” Federal Reserve Bank of Dallas (Sept. 22, 2026)
- U.S. Census Bureau, “Graduating into Disruption: Labor Market Outcomes for AI-Exposed College Majors,” CES-WP-26-56 (Sept. 2026)
- ThePrint, “AI exposure is already affecting college graduates’ first jobs and pay, US study finds” (Sept. 29, 2026)
Leave a Reply