College Majors AI Is Least Likely to Replace (2026 Ranking)

Every parent we talk to asks some version of the same question now: Will this degree still be worth anything in ten years? It's a fair worry. The honest answer is that nobody knows exactly what AI will do to the job market, and anyone who tells you otherwise is selling something.

What we can do is look at the best published research on which jobs AI can actually help do, task by task, and work backward to the majors that lead to those jobs. That's what this ranking is. It doesn't predict layoffs. It shows where the work is most and least exposed to the AI tools that exist today. For a broader, judgment-based list, see our guide to AI-proof college majors.

How did we measure which majors AI is least likely to replace?

The exposure score comes from GPTs are GPTs by Tyna Eloundou, Sam Manning, Pamela Mishkin and Daniel Rock (OpenAI, OpenResearch and the University of Pennsylvania, 2023). Researchers rated every task in the U.S. Department of Labor's O*NET database by whether a large language model, or software built on one, could cut the time to do it by at least half. We use the paper's human-rated "β" score from the published occupation-level data. It runs from 0 (no tasks exposed) to 1 (every task exposed). Lower means less exposed.

The match: for each major we picked the signature occupation most graduates aim for. That mapping is our editorial judgment, and real grads go many places. Pay is the May 2024 national median from the BLS Occupational Employment and Wage Statistics survey. Underemployment for recent grads is from the New York Fed (Feb. 2026 release).

What exposure means: exposure is about tasks, not jobs. A nurse's charting can be sped up by AI; starting an IV can't. The study's authors and Pew Research Center both say directly that exposure doesn't tell you whether jobs will be lost or gained.

Which 13 majors is AI least likely to replace?

Sources: AI exposure = human-rated β from Eloundou et al., GPTs are GPTs (2023), occupation-level data; pay = BLS OEWS May 2024 national median annual wage; underemployment = New York Fed, Feb. 2026 release (recent grads 22–27). *Closest New York Fed category: Medical Technicians (radiologic science), Social Services (social work), Construction Services (construction management). Ranked on unrounded scores. Major-to-job mapping by MajorMatch.

Click any major in the table to open its free loan-vs-upside calculator, or compare debt and lifetime earnings with your own numbers in the degree loan vs. upside calculator. Then take the assessment to see which of these majors actually fit your teen.

What makes each of these majors hard for AI to replace?

Notice the pattern: almost every job near the top involves a body, a place or a person who has to trust you face to face.

1. Aviation (pilots, 0.23)

The highest-paying job on this list by far, at a $226,600 median for airline pilots. Pilots already work alongside heavy automation, and the job is defined by being the accountable human in the cockpit. Air traffic controllers score 0.31.

2. Criminal Justice (police, 0.23)

Patrol work is physical, unpredictable and legally accountable in a way software isn't. The catch is the job market for the degree itself: the New York Fed shows 65.8% of recent criminal justice grads underemployed, the highest of any major. Detectives, who need more investigative judgment, score 0.30 and earn a median of $93,580.

3. Agriculture (farm and ranch managers, 0.24)

Weather, animals, equipment and land don't fit in a chat window. Agricultural managers earn a median of $87,980, though the New York Fed shows 57.1% underemployment for recent agriculture grads, many of whom work in family or field roles.

4. Radiologic Science (rad techs, 0.28)

Positioning a patient, operating imaging equipment and keeping someone calm inside an MRI tube are hands-on tasks. AI reads more images every year, which shifts some work toward radiologists, but the technologist still has to get the image.

5. Social Work (0.30)

Home visits, crisis calls and court appearances are human work. Paperwork is exposed, and many social workers would welcome the help. Pay is the lowest on this list at $58,570, which is worth weighing honestly.

6. Elementary Education (0.31)

Twenty-five eight-year-olds need a person in the room. AI can draft lesson plans, but it can't run recess. The New York Fed shows 16.2% underemployment for elementary education grads, one of the lowest rates of any major.

7. Exercise Science (physical therapists, 0.34)

Physical therapy is hands-on and pays well ($101,020 median), but it requires a doctorate (DPT) after the bachelor's. Exercise physiologists, who don't need one, also score 0.34 but earn a median of $58,160.

8. Secondary Education (0.34)

Slightly more exposed than elementary teaching, since high school work involves more grading and written content, but still firmly in-person.

9. Mechanical Engineering (0.37)

Designing and testing physical things keeps engineers grounded in the real world. AI will speed up the calculations and drafting, not the responsibility for whether the part fails.

10. Civil Engineering (0.38)

Roads, bridges and water systems need licensed engineers to sign off. Civil engineering grads have 15.6% underemployment and a $99,590 median.

11. Nursing (RNs, 0.38)

Charting and care plans are exposed; bedside care isn't. With 3.28 million registered nurses (the largest occupation on this list) and only 12.8% underemployment for nursing grads, it's the safest combination of low exposure and strong demand here. Nurse practitioners score 0.31.

12. Communication Disorders (speech-language pathologists, 0.39)

Working with a child who stutters or a stroke survivor relearning words is personal, clinical work. It requires a master's degree, and the $95,410 median reflects that.

13. Construction Management (0.40)

Running a job site means people, weather, inspections and subcontractors. It's also one of the best majors on our return on investment ranking.

Which majors are most exposed to AI?

The contrast rows in the table tell the story. Writers and authors score 0.81, statisticians 0.71, journalists 0.63, and accountants 0.52. Software developers sit at 0.45 on the human-rated measure. These are jobs built around reading, writing, analyzing and producing text or code, exactly what today's language models do well.

That doesn't make these majors bad bets. Statisticians earn a $103,300 median, and accounting grads have low underemployment. It means the day-to-day work is more likely to change, and graduates will need to be the person who uses the tools well. Pew found something similar: in 2022, 27% of workers with a bachelor's degree or more were in the most AI-exposed jobs, versus 12% of those with only a high school diploma, and the most exposed jobs paid more on average ($33 an hour vs. $20).

Will AI actually replace these jobs?

Nobody can say for sure, and we won't pretend to. The GPTs are GPTs authors estimated that about 80% of U.S. workers could have at least 10% of their tasks affected by large language models, and about 19% could see at least half their tasks affected. That's a big change to how work gets done. Whether it means fewer jobs, different jobs or more productive workers is a separate question the research doesn't answer.

What the data does support: majors that lead to hands-on, licensed, in-person or safety-critical work are less exposed today. That's why so many of them are in healthcare, education, engineering and the trades. If your teen is drawn to those fields, AI is one more reason to feel good about it. If they're drawn to writing or analysis, AI is a reason to pair the major with practical, people-facing skills, not to abandon it.

Are hard majors safer from AI?

Not automatically. Some of the hardest majors on our difficulty ranking, like mechanical and civil engineering, are low-exposure. Others, like mathematics, lead to high-exposure analyst work. And some of the least exposed paths, like elementary education or social work, are not considered hard majors at all. Exposure follows the type of work, not the difficulty of the degree.

How should AI risk factor into choosing a major?

As one input, not the whole decision. A low-exposure major your teen hates is still a bad fit. Start with strengths and interests, then use AI exposure as a tiebreaker between majors that both fit.

Our free assessment does the first part. It matches your teen to 100 majors using our own proprietary scoring method, then you can compare the top picks here and in our AI career risk guide. Run the leading choice through the loan vs. upside calculator to see whether the cost makes sense.

Frequently asked questions