The AI Paradox: The Highest-Paying College Majors Are the Most Likely to Be Automated

Published March 5, 2026 · Data from 24,479 degree programs at 1,890 schools

We analyzed 24,479 bachelor's degree programs across 1,890 schools using three independent AI exposure models. The result was not what we expected.

The college majors most threatened by artificial intelligence currently pay the best. Programs rated "Very High" AI risk earn $51,729 per year in year one. Programs rated "Low" AI risk earn $36,159. That's a 43% gap, and it widens over time: by year five, Very High risk programs pay $75,869 versus $50,068 for Low risk.

So if you're picking a major in 2026, which bet do you make? The degree that pays more now but faces a bigger AI target? Or the safer path that starts lower and may stay there?

24,479

Programs Analyzed

84%

High or Very High Risk

43%

AI Pay Premium

2.2%

Low AI Risk

Finding 1: The AI Earnings Paradox

Higher AI exposure correlates with higher pay. Not sometimes. At every single level.

AI Risk Level	Programs	Avg Year 1 Earnings	Avg Year 5 Earnings	Avg ROI
Very High	7,072	$51,729	$75,869	11.9x
High	13,516	$42,906	$61,047	9.9x
Moderate	3,353	$37,995	$53,248	9.3x
Low	538	$36,159	$50,068	9.2x

Key finding: Very High AI-risk majors earn $15,570 more per year than Low-risk majors in their first year out of school, and $25,801 more by year five. They also have better ROI (11.9x vs 9.2x). The skills that AI threatens are, for now, the same skills the labor market values most.

Why? Because AI targets cognitive work, and cognitive work pays well. Financial analysts, software developers, accountants, data scientists: these are knowledge jobs that command premium salaries. The jobs AI can't touch are the ones where you need to physically be somewhere, doing something with your hands.

Under our pessimistic AI scenario, Very High risk programs face a potential 44% earnings haircut over ten years, worth roughly $33,361 annually. That's enough to erase most of the premium. High risk programs face a 39% haircut ($23,660). Moderate and Low risk programs lose 36% ($19,093) and 27% ($13,553) respectively.

Finding 2: 84% of College Programs Face High or Very High AI Risk

More than four in five degree programs fall into the top two risk categories.

Very High risk: 7,072 programs (28.9%) -- most tasks in the mapped occupations can already be done or heavily assisted by AI
High risk: 13,516 programs (55.2%) -- major task overlap with what AI can do today
Moderate risk: 3,353 programs (13.7%) -- some overlap, but the core work still needs a human in the room
Low risk: 538 programs (2.2%) -- very little overlap with current AI

Only 538 programs out of 24,479 are rated Low risk. That's 2.2%. If you're choosing a major specifically for AI resilience, your options are narrow: primarily healthcare, physical education, skilled trades, and a handful of applied science fields.

Finding 3: The 20 Most AI-Exposed Majors

These are the fields where graduates walk into jobs that overlap the most with what AI can already do.

Major	Programs	AI Risk Score	Avg Year 1 Earnings	Risk Level
Ethnic, Cultural & Gender Studies	92	0.671	$32,229	Very High
Literature	8	0.667	$23,375	Very High
Philosophy & Religious Studies	77	0.653	$32,298	Very High
Management Information Systems	153	0.647	$59,612	Very High
Communication & Journalism	16	0.645	$34,532	Very High
Computer Programming	31	0.641	$59,150	Very High
Statistics	51	0.639	$63,187	Very High
English Language & Literature	493	0.631	$30,428	Very High
Accounting & Related Services	714	0.630	$53,725	Very High
Mathematics	253	0.626	$50,797	Very High
Computer Engineering	174	0.621	$78,695	Very High
Computer Science	345	0.609	$73,714	Very High
Journalism	178	0.610	$33,851	Very High
Economics	336	0.602	$54,038	Very High
Finance & Financial Management	409	0.598	$57,962	Very High
Political Science & Government	395	0.594	$37,155	Very High
History	325	0.587	$32,741	Very High
Business Admin & Management	1,187	0.580	$45,884	Very High
Marketing	357	0.573	$44,137	Very High
Psychology	854	0.561	$32,155	Very High

Accounting is the biggest single major in the Very High risk zone: 714 programs. That's a lot of students walking into a field where AI already handles a chunk of the daily work. Business Administration (1,187 programs) and Psychology (854 programs) are even larger, though they score slightly lower.

There's also a huge earnings split within Very High. Computer Engineering grads earn $78,695. English majors earn $30,428. Their AI risk scores are almost identical (0.621 vs 0.631). AI exposure alone doesn't tell you what you'll earn. It depends on what kind of AI-exposed work you're doing.

Finding 4: The Most AI-Resistant Majors

The majors with the lowest AI exposure all have something in common: you can't do them from behind a screen.

Major	Programs	AI Risk Score	Avg Year 1 Earnings	Risk Level
Liberal Arts & General Studies	424	0.160	$37,173	Low
Marine Transportation	7	0.176	$82,678	Low
Dance	62	0.213	$22,513	Low
Fishing & Fisheries Sciences	10	0.238	$33,061	Low
Funeral Service & Mortuary Science	6	0.243	$48,871	Low
Culinary Arts & Food Service	12	0.243	$28,611	Low
Practical & Vocational Nursing	6	0.245	$13,597	Low
Agricultural Mechanization	10	0.250	$56,147	Moderate
Health & Physical Education	529	0.277	$31,057	Moderate
Clinical & Medical Lab Science	99	0.286	$65,728	Moderate
Dental Support Services	74	0.314	$60,938	Moderate
Allied Health Professions	195	0.325	$59,453	Moderate
Registered Nursing	941	0.427	$75,165	High

The jobs AI can't do are the ones that require a body. Marine captains, nurses, dental hygienists, morticians, lab techs, gym teachers. You need to physically show up. No chatbot is embalming a corpse or steering a cargo ship.

Marine Transportation is the sleeper here. AI risk of 0.176 and earnings of $82,678. Only 7 programs offer it, but if you can get in, it's one of the few majors that scores well on both safety and pay. Clinical & Medical Lab Science and Allied Health offer something similar at a bigger scale.

Then there's Registered Nursing: 941 programs, $75,165 average earnings, risk score of 0.427. It technically lands in the High category, but that's well below computer science (0.609) or accounting (0.630). Nursing pays more, has more openings, and is harder for AI to replace. It might be the single best risk-adjusted bet in higher education right now.

Finding 5: How AI Changes the 10-Year Earnings Picture

Current earnings are backward-looking. What matters for an 18-year-old picking a major is what happens over the next decade. We modeled three scenarios for selected programs.

Computer Science (UC Berkeley, DegreeWorth Score: 78)

Scenario	10-Year Cumulative Earnings	Impact
Optimistic (no disruption)	$2,471,342	—
Base Case (gradual AI adoption)	$1,649,244	-33%
Pessimistic (aggressive displacement)	$895,468	-64%

AI risk score: 0.609 (Very High). Under the pessimistic scenario, only 8 of 13 mapped career paths remain fully viable. Cumulative earnings drop by $1.6 million over a decade.

Accounting (CSU Northridge, DegreeWorth Score: 77)

Scenario	10-Year Cumulative Earnings	Impact
Optimistic	$952,115	—
Base Case	$647,367	-32%
Pessimistic	$427,671	-55%

AI risk score: 0.630 (Very High). Under the pessimistic scenario, only 10 of 14 career paths remain fully viable.

Registered Nursing (CSU Stanislaus, DegreeWorth Score: 81)

Scenario	10-Year Cumulative Earnings	Impact
Optimistic	$1,233,066	—
Base Case	$1,188,392	-4%
Pessimistic	$953,824	-23%

AI risk score: 0.427 (High). Under the pessimistic scenario, all 6 of 6 career paths remain viable. Even aggressive AI adoption barely dents nursing earnings.

English (UH Clear Lake, DegreeWorth Score: 54)

Scenario	10-Year Cumulative Earnings	Impact
Optimistic	$579,926	—
Base Case	$448,634	-23%
Pessimistic	$334,565	-42%

AI risk score: 0.631 (Very High). Under the pessimistic scenario, only 2 of 3 career paths remain viable.

Key finding: Nursing earns more than English in every scenario and loses the least to AI disruption. Under the pessimistic scenario, a nursing grad from CSU Stanislaus earns nearly 3x what an English grad from UH Clear Lake earns ($953K vs $335K over 10 years), despite nursing facing a nominally "High" risk label. The risk label matters less than the magnitude of impact.

Finding 6: Two Different Kinds of AI Exposure

Our composite risk score blends three models, and they disagree in interesting ways. The type of AI exposure matters as much as the amount.

GPT task exposure (OpenAI's research) measures what percentage of a job's tasks LLMs can perform or speed up. AIOE scores (Felten et al.) measure broader overlap with AI capabilities like image recognition, translation, and reasoning. They're measuring different things, and the gap between them tells a story.

Major	GPT Exposure	AIOE Score	Pattern
Computer Programming	0.777	0.183	High GPT, Low AIOE
Computer Science	0.724	0.180	High GPT, Low AIOE
Computer Engineering	0.705	0.476	High GPT, Moderate AIOE
Philosophy	0.638	1.376	Moderate GPT, Very High AIOE
Statistics	0.638	1.209	Moderate GPT, Very High AIOE
Accounting	0.616	1.252	Moderate GPT, Very High AIOE
English	0.612	1.299	Moderate GPT, Very High AIOE
Journalism	0.644	0.802	Moderate GPT, High AIOE

Computer fields score very high on GPT exposure but low on AIOE. LLMs can write code, debug, and document, but broader AI systems don't overlap much with what programmers actually do all day. The likely outcome: the job changes shape. You go from writing code to directing AI that writes code. The role survives, but it looks different.

Humanities are the opposite. Moderate GPT exposure, very high AIOE. Writing, analysis, research, translation: AI overlaps with the whole skill set, not just specific tasks. A programmer can redefine their role around AI. It's harder to see how a philosophy grad does the same thing.

Full Major Rankings: AI Risk for Every Field of Study

The complete table below includes every major with at least 5 programs in our dataset. Click any column header to sort. Use the search box to find a specific field.

Major ▲	Programs ▲	AI Risk ▲	Avg Year 1 ▲	Avg Year 5 ▲	Risk Level ▲
Ethnic, Cultural & Gender Studies	92	0.671	$32,229	$51,769	Very High
Literature	8	0.667	$23,375	$49,856	Very High
Philosophy & Religious Studies	77	0.653	$32,298	$55,026	Very High
Management Information Systems	153	0.647	$59,612	$83,951	Very High
Communication & Journalism	16	0.645	$34,532	$55,432	Very High
Computer Programming	31	0.641	$59,150	$83,390	Very High
Statistics	51	0.639	$63,187	$99,749	Very High
English Language & Literature	493	0.631	$30,428	$48,551	Very High
Accounting & Related Services	714	0.630	$53,725	$72,384	Very High
Mathematics	253	0.626	$50,797	$76,592	Very High
Computer Engineering	174	0.621	$78,695	$110,795	Very High
Computer Science	345	0.609	$73,714	$112,556	Very High
Journalism	178	0.610	$33,851	$52,417	Very High
Economics	336	0.602	$54,038	$82,645	Very High
Finance & Financial Management	409	0.598	$57,962	$84,751	Very High
Political Science & Government	395	0.594	$37,155	$59,847	Very High
History	325	0.587	$32,741	$53,182	Very High
Business Admin & Management	1,187	0.580	$45,884	$64,503	Very High
Marketing	357	0.573	$44,137	$63,891	Very High
Psychology	854	0.561	$32,155	$47,832	Very High
Sociology	348	0.555	$31,847	$47,115	Very High
Mechanical Engineering	303	0.548	$71,284	$96,438	Very High
Chemical Engineering	145	0.541	$72,659	$98,214	Very High
Electrical Engineering	261	0.537	$77,489	$107,326	Very High
Biology	645	0.520	$33,741	$51,287	High
Civil Engineering	196	0.512	$63,847	$86,159	High
Criminal Justice	505	0.498	$37,284	$52,619	High
Social Work	337	0.482	$35,971	$46,284	High
Education	732	0.465	$36,547	$47,891	High
Registered Nursing	941	0.427	$75,165	$94,738	High
Allied Health Professions	195	0.325	$59,453	$67,855	Moderate
Dental Support Services	74	0.314	$60,938	$63,145	Moderate
Clinical & Medical Lab Science	99	0.286	$65,728	$73,642	Moderate
Health & Physical Education	529	0.277	$31,057	$53,626	Moderate
Agricultural Mechanization	10	0.250	$56,147	$70,812	Moderate
Funeral Service & Mortuary Science	6	0.243	$48,871	$57,989	Low
Culinary Arts & Food Service	12	0.243	$28,611	$38,104	Low
Fishing & Fisheries Sciences	10	0.238	$33,061	$46,598	Low
Dance	62	0.213	$22,513	$34,024	Low
Marine Transportation	7	0.176	$82,678	$111,365	Low
Liberal Arts & General Studies	424	0.160	$37,173	$49,958	Low

Showing majors with 5 or more programs. Sorted by AI risk score (highest first). Click column headers to re-sort.

Look Up Your Specific School + Major

AI risk varies by school. See the detailed AI scenario analysis for your exact program.

Explore Degree Values

Methodology

This analysis covers 24,479 bachelor's degree programs across 1,890 schools in the United States.

Data Sources

Earnings & debt data: U.S. Department of Education College Scorecard, field-of-study level data (2024 release). Reports actual median earnings 1 and 4 years after graduation.
Tuition data: NCES Integrated Postsecondary Education Data System (IPEDS), institutional characteristics survey.
Job growth projections: Bureau of Labor Statistics Employment Projections, 2024–2034.
Occupation mapping: NCES CIP-to-SOC crosswalk, linking 6-digit academic program codes to 6-digit occupation codes.

AI Risk Models

We compute a composite AI risk score from three independent, peer-reviewed or institutionally published models:

GPT Task Exposure (Eloundou et al., OpenAI, 2023) — Measures the percentage of occupational tasks that can be performed or substantially accelerated by large language models. Score range: 0 to 1.
AI Occupational Exposure (AIOE) (Felten et al., 2023) — Measures overlap between AI capabilities (image recognition, language modeling, translation, reasoning) and occupational task requirements. Normalized score.
Automation Probability (Frey & Osborne, 2017) — Classic estimate of the probability that an occupation can be automated, based on required manual dexterity, creativity, and social intelligence. Score range: 0 to 1.

Composite Score Formula

The composite AI risk score for each occupation is:

AI Risk = 0.4 × GPT Exposure + 0.3 × AIOE (normalized) + 0.3 × Frey-Osborne Probability

The discount applied to earnings projections is capped at 0.50 (50%) to avoid projecting total job elimination, which no credible model predicts.

For programs that map to multiple occupations via the CIP-SOC crosswalk, we compute a weighted average based on the occupation's share of employment among graduates of that program.

Risk Categories

Very High (score ≥ 0.50): A majority of mapped occupational tasks overlap with current AI capabilities
High (score 0.35–0.49): Significant task overlap with AI capabilities
Moderate (score 0.25–0.34): Some task overlap; core functions require human judgment or physical presence
Low (score < 0.25): Minimal overlap with current AI capabilities

Three-Scenario Modeling

For each program, we project 10-year cumulative earnings under three scenarios:

Optimistic: No AI disruption. Earnings follow current BLS growth projections.
Base Case: Gradual AI adoption. Earnings are reduced by the composite AI risk score, phased in over 10 years.
Pessimistic: Aggressive AI displacement. Full AI risk discount applied, with occupations below a viability threshold excluded from earnings projections entirely.

Limitations

College Scorecard earnings data reflects outcomes 1 and 4 years post-graduation. Long-term career trajectories (10+ years) are projected, not observed.
AI exposure models were published between 2017 and 2023. AI capabilities are evolving faster than these models anticipated; actual disruption may be higher or lower than modeled.
The CIP-SOC crosswalk maps degree programs to likely occupations, but individual career paths vary. A philosophy major who becomes a software engineer won't match the predicted outcome for philosophy.
Programs with suppressed earnings data (fewer than 10–20 graduates reporting) are excluded from this analysis.

Cite This Report

DegreeWorth. "The AI Paradox: The Highest-Paying College Majors Are the Most Likely to Be Automated." DegreeWorth Reports, March 5, 2026. https://degreeworth.com/blog/ai-risk-college-majors.html