Degrees in danger? These college majors face the biggest AI job shock

A US Census working paper found that graduates from the most AI-exposed college majors (subjects) saw initial employment fall by five percentage points and first full-quarter earnings by 13 percent after ChatGPT, with computer science and related fields among the most exposed. Check the full list of AI-exposed college majors divided into 10 levels.

In Short

• US Census data linked ChatGPT-era changes to weaker first-job outcomes for AI-exposed fields
• Initial earnings fell 13% for graduates in the most AI-exposed 10% of college majors
• The US findings offer an early warning for Indian students choosing AI-exposed degrees

For years, students were told that a degree in computer science, engineering or another technical field was a strong route into the job market. A new US Census Bureau working paper suggests that the first step into those careers may now be getting harder.

The paper, 'Graduating into Disruption: Labor Market Outcomes for AI-Exposed College Majors', tracked administrative records covering about 6.67 million bachelor's degree graduates, roughly 29% of US bachelor's degrees awarded between 2016 and 2024.

Researchers compared outcomes across college majors according to how exposed the occupations associated with those majors were to AI.

The key break came after ChatGPT became widely available in late 2022. Among the 10% of majors with the highest AI exposure, the probability of being employed shortly after graduation fell by five percentage points, while full-quarter initial earnings fell by 13% relative to less-exposed fields.

The researchers said the earnings decline was comparable in magnitude to the losses associated with graduating during a large recession.

WHICH DEGREES ARE MOST AI-EXPOSED?

Importantly, the study does not produce a simple list of "safe" and "unsafe" degrees. It divides majors into 10 exposure deciles using an AI exposure score based on the occupations graduates typically entered before ChatGPT.

The top decile is heavily concentrated in computing and software-related fields. Journalism also appears in this highest exposure group.

The ranking is based on the paper's GPT-4 Beta exposure measure, not on whether a particular profession will disappear.

INITIAL EARNINGS DROP BY 13%: EXPLAINED

The 13% figure is not a 13% fall in the number of jobs, and it is not a prediction that computer science graduates will earn 13% less throughout their careers.

It refers to full-quarter initial earnings among graduates in the most AI-exposed decile. The study found the earnings gap became smaller as graduates moved further into their careers, falling to about 5% after two years. But the researchers say their data are not long enough to establish whether these graduates eventually catch up with earlier cohorts.

There is another revealing part of the result. Roughly half of the earnings decline came from lower earnings within industries. The other half came from graduates shifting towards lower-paying sectors, including retail and accommodation and food services.

In other words, the issue is not as simple as AI cutting the salary of the job you studied for. For some graduates, the bigger problem may be getting that first job in the field at all.

THIS IS NOT PROOF THAT AI CAUSED EVERY JOB LOSS

The timing is notable: the divergence in employment, earnings and job switching began around the introduction of ChatGPT. But the researchers do not claim that every change was directly caused by companies replacing workers with AI.

They tested alternative explanations including graduate supply, work-from-home intensity, self-employment and postgraduate study. Their results were not readily explained by a simple increase in the number of graduates. However, they also acknowledge that the short time series means the long-term effects remain uncertain.

And there is an important publishing caveat: this is a Census Bureau working paper, not a peer-reviewed Census publication. The paper itself states that CES working papers have not undergone the review accorded to Census Bureau publications and that its conclusions belong to the authors, not the agency.

WHAT DOES THIS MEAN FOR INDIA?

The US findings cannot simply be pasted onto Indian degrees.

India's labour market, college system, hiring patterns and degree-to-job pathways are different. But the question behind the research is very familiar to Indian students and parents: what happens when the first rung of a career ladder starts shrinking?

There are already signs of a rapidly changing Indian market. Foundit's annual tracker recorded 2,90,256 AI-linked job postings in India in 2025, with AI hiring projected to rise further in 2026.

At the same time, a LinkedIn survey reported that 84% of Indian professionals felt unprepared to find a new job in 2026, while 72% said they were actively looking for a new role.

Those numbers do not establish that Indian graduates face the same degree-level shock found in the US study. But they do show why the question is becoming harder to ignore.

For students choosing a degree in 2026, the useful question may no longer be simply "Will AI replace this career?" It may be: "Which parts of this career will AI change, and what skills will still get me through the first door?"

- Ends