StableJob
AI exposure by college major · Evidence review

What the research says about AI and graduate jobs

Researchers disagree about whether AI is hurting recent graduates. The answer depends on the data, the outcome measured and the group studied.

Below are five sources and what each can and cannot show. Our major pages do not use these studies to score a degree. They score the jobs a degree leads to, using Census data on where graduates work. The studies are background so you can judge how much weight to give the scores.

Studies that find worse outcomes in AI-exposed fields

Census Bureau working paper

Orr, Tucker and Warren (CES 26-56, September 2026) link administrative records on graduates, covering about 29% of US bachelor's degrees awarded from 2016 to 2024, to employer data. They rank majors by how exposed to AI the occupations of their graduates are. Graduates in the most exposed tenth of majors were five percentage points less likely to be employed in the quarter after graduation, and their first full-quarter earnings were about 13% lower, easing to about 5% after two years, compared with graduates of the least exposed six tenths of majors.

The most exposed group is dominated by computer science, information systems and related software degrees. The authors conclude that generative AI played a substantial role. They also find that more graduates in that group went to graduate school or into self-employment, which, taken at face value, could explain up to half of the five-point employment gap. The paper has not been through the Census Bureau's publication review, and its views are the authors', not the Bureau's.

Federal Reserve Bank of Dallas

Dodini and Smith (22 September 2026) study graduates of Texas public universities only. They rate majors by how much of their typical jobs' tasks generative AI could do. A 10-point higher exposure was associated with a 1.7-point lower chance of finding Texas employment within a year, and about 5% lower first-year earnings among those who found work. The authors conclude that generative AI caused these declines. The data cover Texas public universities and Texas jobs only, so the results may not apply elsewhere.

Studies that do not find it

NBER Working Paper 35796

Fairlie and Wu (September 2026) use national survey data for summer 2026. They find that recent-graduate unemployment did not spike compared with earlier summers, older graduates, or young workers without a degree. They report some evidence of a link with how much of a job can be done remotely.

Economic Policy Institute

Gould and Fast (21 May 2026) find that 85% of young graduates work in occupations with strong recent job growth. Graduates are more likely than other workers to be in AI-exposed occupations, but young workers without degrees also saw unemployment rise over three years, which EPI reads as a sign that AI is unlikely to be the main driver. EPI also cites other work, using payroll data, that finds larger declines for young workers in AI-exposed occupations.

Employ America (older)

Raderman (22 July 2025) is older than the studies above. He found no consistent link between a major's AI exposure and its unemployment rate: mathematics, accounting and business analytics had lower unemployment than before the pandemic, while computer science, computer engineering and chemistry had higher.

How to read this

  • The studies measure different things: employment and earnings in administrative records, or unemployment in surveys.
  • They cover different groups: one state, or the whole country, and different years.
  • They use different ways of deciding which majors are "exposed".
  • These are comparisons between groups, not experiments. Some authors conclude AI played a role and others find no effect. None can rule out every other cause.
  • Employ America reports that unemployment among graduates aged 22 to 27 has been above the overall rate since late 2018, before generative AI was widely used.

A fair summary: software-related degrees are the most scrutinised and the main place where harm is reported. Other fields have far less evidence either way, and the picture may change as AI use grows. If you are choosing a degree, look at where its graduates actually work, which is what each major page on this site shows, rather than relying on a single headline.

How this looks for specific degrees

Each degree page applies this background to one field and shows where its graduates actually work. A few to start with:

Sources

Last reviewed 2026-10-27. Next review 2027-04-27. This page is general information, not career or educational advice.