The 25 Most AI-Resistant Careers of 2026: A DreamJobMatcher Analysis
We scored 800+ careers on automation risk using public labor data and our proprietary AI Stability framework.
An AI-resistant career is one where the core daily work cannot be decomposed into discrete, automatable tasks because it depends on physical presence, complex human judgment, regulated trust, or high-context interpersonal interaction. AI-resistant does not mean AI-immune. It means the role remains net additive when AI is introduced, with productivity gains rather than headcount displacement.
This study ranks the 25 most AI-resistant careers in the United States for 2026 by combining U.S. Bureau of Labor Statistics 2024 to 2034 occupational projections with DreamJobMatcher’s proprietary AI Stability Score, a 0 to 100 measurement of automation exposure across four dimensions. We scored 800+ occupations in the BLS Occupational Outlook Handbook and selected the 25 with the highest stability scores that also have BLS-projected employment growth at or above the national average.
The pattern is consistent with peer-reviewed research from the OECD AI and labour markets program and the World Economic Forum Future of Jobs Report 2025: roles that combine human-trust requirements with physical or regulated work consistently outperform on durability, while purely cognitive desk roles that lack judgment density are most exposed. McKinsey’s labor-market work, including the Generative AI and the Future of Work in America report, reaches similar conclusions on the directional shift.
Methodology in one paragraph
The DreamJobMatcher AI Stability Score is a framework estimate, modeled across four dimensions, each weighted 25%: task decomposability (how cleanly the work splits into discrete, repeatable subtasks), judgment load (the proportion of daily decisions requiring contextual, ethical, or experiential reasoning), human interaction density (the share of work time spent in real-time interpersonal exchange), and regulatory complexity (the degree to which the role is licensed, audited, or legally bounded). Inputs include O*NET task taxonomies, BLS employment projections, and a curated 2024 to 2026 review of AI deployment cases in each occupation. The four-dimensional approach is documented in our PATH method reference and applied across our career assessment.
The 25 most AI-resistant careers of 2026
Scores range from 78 to 95. Median salary figures are 2023 BLS data, the most recent full-year set published. Holland Code reflects the dominant RIASEC types per O*NET. “Why it’s resistant” summarizes the dominant stability dimensions for that role.
| Rank | Career | AI Stability | Median salary | Holland | Why it’s resistant |
|---|---|---|---|---|---|
| 1 | Registered nurse | 94 | $86,070 | SIR | Physical presence, judgment under uncertainty, regulated practice, patient trust |
| 2 | Mental health counselor | 93 | $53,710 | SAI | Therapeutic alliance, licensure, contextual emotional reasoning |
| 3 | Physical therapist | 92 | $99,710 | SIR | Hands-on assessment, custom progression, doctoral licensure |
| 4 | Surgeon | 92 | $239,200 | IRS | Sterile-field judgment, intraoperative adaptation, malpractice exposure |
| 5 | Primary care physician | 91 | $229,300 | SIA | Diagnostic synthesis across noisy inputs, longitudinal patient relationships |
| 6 | Psychiatrist | 91 | $249,760 | ISA | Clinical interview, prescribing authority, medico-legal risk |
| 7 | Occupational therapist | 90 | $96,370 | SIA | Individualized rehabilitation plans, in-home assessment |
| 8 | Social worker (clinical) | 89 | $58,380 | SAE | Crisis intervention, regulated case management, court testimony |
| 9 | Special education teacher | 89 | $65,910 | SAE | Individualized education plans, behavioral judgment, parent partnerships |
| 10 | Veterinarian | 88 | $119,100 | IRS | Non-verbal patient assessment, surgical work, owner counseling |
| 11 | Firefighter | 88 | $57,120 | RSE | Physical response, real-time risk assessment, civic trust |
| 12 | Paramedic | 87 | $53,180 | RSI | Field triage, time-critical decisions, regulated scope of practice |
| 13 | Electrician | 87 | $61,590 | RIC | Code compliance, on-site diagnosis, physical install |
| 14 | Plumber | 86 | $61,550 | RIC | In-situ problem-solving, licensure, emergency response |
| 15 | HVAC technician | 85 | $57,300 | RIC | System diagnosis on heterogeneous installs, refrigerant certification |
| 16 | Police detective | 85 | $91,090 | SEI | Witness interviewing, courtroom evidence, ethical judgment |
| 17 | Dental hygienist | 84 | $87,530 | SIR | Manual procedure, patient education, regulated practice |
| 18 | Elementary school teacher | 84 | $63,680 | SAE | Classroom management, social-emotional teaching, parent communication |
| 19 | Judge | 84 | $148,030 | ESI | Constitutional interpretation, demeanor assessment, appellate accountability |
| 20 | Conductor (orchestra) | 83 | $62,940 | ASE | Live ensemble interpretation, rehearsal craft, performance presence |
| 21 | Postsecondary professor | 82 | $84,380 | ISA | Original research, peer review, mentorship at scale |
| 22 | Civil engineer | 81 | $95,890 | IRC | Licensed sign-off (PE), site-specific judgment, public-safety liability |
| 23 | Urban and regional planner | 80 | $81,800 | ISE | Stakeholder synthesis, zoning judgment, multi-decade horizon |
| 24 | Art therapist | 79 | $56,560 | SAI | Creative-clinical hybrid, in-person therapeutic relationship |
| 25 | Music therapist | 78 | $55,290 | SAI | Live therapeutic intervention, board certification, individualized treatment |
If the cluster on this list resonates, see how it maps to personality fit in our Holland Code guide and Big Five career guide. Several of these roles also appear in our deeper personality-aligned guides, including INFJ careers for therapeutic roles and ISTP careers for the skilled trades.
Reading the scores. A score of 85 does not mean a role is 85% protected. It means the role exhibits the structural properties that historically correlate with AI augmentation rather than substitution. Even high-scoring roles will see significant workflow changes by 2034, just not headcount collapse.
What makes a career AI-resistant?
The DJM framework isolates four dimensions that drive durability. Most legacy automation indices focus on task decomposability alone. That misses the structural variables that explain why a paralegal contract-review role is exposed while a litigation paralegal who interviews witnesses is not. The four dimensions, in detail:
1. Task decomposability (low is good)
Can the role be split into discrete, repeatable subtasks each of which can be specified, automated, and re-assembled? Tax-form data entry is high decomposability. Family-court mediation is low. The lower the decomposability, the higher the durability. See the OECD analysis of task structure and AI exposure for the underlying mechanism.
2. Judgment load (high is good)
What proportion of daily decisions require contextual reasoning, ethical weighing, or experience-driven intuition that resists explicit specification? A surgeon adjusting mid-operation has high judgment load. A transcriptionist applying punctuation rules has low load. AI systems compress task time, but unspecified judgment remains the human’s job.
3. Human interaction density (high is good)
How much of the work is real-time interpersonal exchange that requires reading emotional cues, building trust, or co-regulating with another human? Therapy, teaching, and skilled-trade customer work all score high. Backend data engineering scores low. The WEF Future of Jobs Report 2025 identifies interpersonal density as one of the top three predictors of durable roles through 2030.
4. Regulatory complexity (high is good)
Is the role licensed, audited, or otherwise legally bounded such that an AI cannot perform the work without a human accountable party? Physicians, lawyers, civil engineers, electricians, and judges all sit behind licensure walls that AI cannot cross unilaterally. Regulated roles do not just have legal moats. They have liability structures that require an accountable human.
Roles that score high on three or four dimensions cluster near the top of our 2026 list. Roles that score high on only one dimension (for example, regulatory complexity for a junior compliance analyst whose actual work is checklist review) sit much lower than their licensure suggests. For a fuller treatment of how we test for these dimensions, see our AI durability guide.
Which industries are most AI-resistant overall?
Aggregating the per-role scores up to industry level using BLS employment weights produces a clear hierarchy. The eight industries below have the highest weighted average AI Stability Score for 2026.
| Industry | Avg AI Stability | Driver dimension | BLS 2024-2034 growth |
|---|---|---|---|
| Healthcare and social assistance | 87 | Human interaction + regulatory | +9% (well above national) |
| Skilled construction trades | 83 | Physical presence + judgment | +5% |
| Public safety and emergency services | 82 | Physical + interpersonal | +4% |
| Education (K-12 + special ed) | 80 | Interpersonal density | +1% (below average but stable) |
| Civil and environmental engineering | 78 | Regulatory + judgment | +6% |
| Skilled installation and repair (HVAC, electrical, plumbing) | 82 | Physical + judgment | +6% to +9% |
| Mental and behavioral health services | 86 | Interpersonal + regulatory | +18% (top decile) |
| Veterinary services | 85 | Physical + judgment + interpersonal | +19% |
The list is notably absent of pure white-collar information work. That absence is the story. Industries that produce, repair, treat, teach, or judge dominate the durability leaderboard. Industries that primarily move information between databases do not. For a parallel look at where these industries also grow fastest, read our companion analysis on the fastest-growing AI-proof industries of 2026.
Which roles are MOST at risk in 2026?
The contrast matters. Without the low-scoring roles, the high-scoring list looks like marketing. The roles below score between 12 and 30 on AI Stability. They share a profile: high task decomposability, low judgment load, low or screen-mediated interpersonal density, and minimal regulatory protection. Many are entry-level roles, which has the secondary effect of cutting traditional ladder rungs for new graduates.
| Role | AI Stability | Why it’s exposed |
|---|---|---|
| Data entry clerk | 12 | Pure structured input, OCR plus LLMs displace baseline work |
| Telemarketer | 15 | Scripted outbound calls, voice AI now matches conversion at lower cost |
| Content moderator (basic) | 18 | Policy lookup against image and text, models classify at platform scale |
| Transcriptionist | 19 | Whisper-class models match human accuracy on clean audio |
| Basic bookkeeper | 22 | Categorization and reconciliation are mostly automated in modern stacks |
| Junior coder (CRUD) | 25 | Code completion plus agentic IDEs collapse routine implementation work |
| Customer service rep (tier 1) | 26 | Knowledge-base answering, return processing, status lookup all automatable |
| Paralegal (contract review) | 27 | Document review and citation work are model-replicable; litigation paralegal not |
| Junior market research analyst | 28 | Survey coding and report drafting collapse with AI tooling |
| Junior copywriter | 30 | Output-heavy, judgment-light writing is most exposed segment |
The pattern is not that these jobs disappear in 2026. It is that the entry-level versions are being absorbed, and the seniors in those fields are being asked to do more with smaller teams. For graduates choosing a path, the implication is sharp: pick fields where the entry-level work still requires the four durability dimensions. For mid-career workers in exposed roles, the move is toward the senior, judgment-heavy, or regulated variant of the field. See our guides on career change in your 30s and skills assessment for career planning.
How the AI Stability Score is calculated
Each occupation’s task list from O*NET is mapped against the four dimensions on a 0 to 100 scale. The four dimension scores are averaged, then adjusted by two modifiers: a labor-market multiplier reflecting BLS 2024 to 2034 projected employment change, and a deployment-evidence modifier reflecting whether public 2024 to 2026 cases show AI being used as augmentation (positive) or substitution (negative) for that role. The final number is the AI Stability Score we publish per role.
We chose four dimensions rather than a single composite because composite indices tend to mask which lever moves a score. A surgeon and a master electrician score similarly overall but for very different reasons: the surgeon scores on judgment and regulation; the electrician scores on physicality and judgment. The four-dimensional view lets a reader, a journalist, or a policymaker see the mechanism, not just the rank. The full methodology is documented at our PATH method page, and we apply the framework continuously in our DJM career assessment. For peer-reviewed context, see McKinsey’s Generative AI and the Future of Work in America and the OECD’s ongoing labour-market AI program.
What this means if you’re choosing a career in 2026
Three implications follow from the data. First, fields that combine human judgment with regulated practice and physical or interpersonal presence are the safest long-horizon bet. That covers most of healthcare, the skilled trades, education, public safety, and licensed professional services. Second, exposure inside a field varies by role variant. A junior copywriter is at risk; a senior brand strategist directing AI-generated drafts is not. The same logic applies in legal, finance, software, and design. Third, the AI-leveraged version of any role outperforms the unleveraged version. The right strategy is rarely to flee AI. It is to choose a role where humans remain accountable and to learn to direct AI tools inside that role.
Practically, the move is to take a structured assessment that ranks roles by fit and by AI Stability together, then build skills inside the surviving cluster. The DreamJobMatcher assessment scores all 800+ occupations against your Holland Code, your Big Five profile, and the AI Stability framework above, then returns your top 25 matches with each role’s durability score attached. Browse adjacent reading: INTJ careers, ISFJ careers, our full career guide, and best career tests of 2026.
Frequently asked questions
Are any careers truly AI-proof?
No career is fully AI-proof. The 25 roles in this study are AI-resistant, meaning the structural properties of the work make substitution unlikely on a 10-year horizon. Workflow change inside the role is near-certain. The right framing is durability, not immunity.
How often is the AI Stability Score updated?
The score is reviewed quarterly. The four dimension weights are stable, but the deployment-evidence modifier shifts as new 2024 to 2026 case studies become public. The 2026 ranking reflects evidence through Q1 2026.
Does a high AI Stability Score guarantee good pay or growth?
No. Stability and growth are separate axes. Our companion analysis on the fastest-growing AI-proof industries of 2026 cross-references the two so readers can find roles that are both durable and growing.
How does DJM’s framework differ from older automation indices?
Earlier indices (Frey and Osborne 2013, OECD task-based 2018) measured task decomposability alone. The DJM framework adds judgment load, interpersonal density, and regulatory complexity. The result is a fuller picture that explains why, for example, paralegals split into exposed (contract review) and resistant (litigation) tracks.
Should a 22-year-old avoid all roles with low AI Stability?
Not necessarily. Some low-score roles are still on-ramps to high-score senior roles. The risk is treating the low-score role as a 30-year path. The right strategy is to enter with a clear 3-to-5-year transition plan into a higher-durability variant of the field.
Press inquiries
Journalists and editors covering AI, the labor market, or the future of work are welcome to cite this study and the AI Stability Score methodology. Contact [email protected] for data extracts, expert commentary, or interview requests.
{“@context”:”https://schema.org”,”@graph”:[
{“@type”:”NewsArticle”,”headline”:”The 25 Most AI-Resistant Careers of 2026: A DreamJobMatcher Analysis”,”description”:”DreamJobMatcher ranks the 25 most AI-resistant careers in 2026 using BLS data and the proprietary AI Stability Score across four dimensions.”,”datePublished”:”2026-05-22″,”dateModified”:”2026-05-22″,”author”:{“@type”:”Organization”,”name”:”DreamJobMatcher Research Team”},”publisher”:{“@type”:”Organization”,”name”:”DreamJobMatcher”,”url”:”https://dreamjobmatcher.com”},”mainEntityOfPage”:”https://dreamjobmatcher.com/most-ai-resistant-careers-2026/”},
{“@type”:”BreadcrumbList”,”itemListElement”:[{“@type”:”ListItem”,”position”:1,”name”:”Home”,”item”:”https://dreamjobmatcher.com/”},{“@type”:”ListItem”,”position”:2,”name”:”Industry & Future of Work”,”item”:”https://dreamjobmatcher.com/category/industry-future-of-work/”},{“@type”:”ListItem”,”position”:3,”name”:”25 Most AI-Resistant Careers of 2026″,”item”:”https://dreamjobmatcher.com/most-ai-resistant-careers-2026/”}]},
{“@type”:”Dataset”,”name”:”DreamJobMatcher AI Stability Score 2026″,”description”:”AI Stability Scores for 25 occupations, 0-100 scale, derived from O*NET task taxonomies, BLS 2024-2034 projections, and DJM’s four-dimensional framework.”,”creator”:{“@type”:”Organization”,”name”:”DreamJobMatcher”},”license”:”https://creativecommons.org/licenses/by/4.0/”,”url”:”https://dreamjobmatcher.com/most-ai-resistant-careers-2026/”,”keywords”:”AI stability, automation risk, BLS, occupations, 2026″},
{“@type”:”FAQPage”,”mainEntity”:[
{“@type”:”Question”,”name”:”Are any careers truly AI-proof?”,”acceptedAnswer”:{“@type”:”Answer”,”text”:”No career is fully AI-proof. The 25 roles in this study are AI-resistant, meaning the structural properties of the work make substitution unlikely on a 10-year horizon.”}},
{“@type”:”Question”,”name”:”How often is the AI Stability Score updated?”,”acceptedAnswer”:{“@type”:”Answer”,”text”:”The score is reviewed quarterly. Dimension weights are stable, but the deployment-evidence modifier shifts as new 2024-2026 case studies become public.”}},
{“@type”:”Question”,”name”:”Does a high AI Stability Score guarantee good pay or growth?”,”acceptedAnswer”:{“@type”:”Answer”,”text”:”No. Stability and growth are separate axes. See our companion analysis on the fastest-growing AI-proof industries of 2026 to cross-reference.”}},
{“@type”:”Question”,”name”:”How does DJM’s framework differ from older automation indices?”,”acceptedAnswer”:{“@type”:”Answer”,”text”:”Earlier indices measured task decomposability alone. The DJM framework adds judgment load, interpersonal density, and regulatory complexity.”}},
{“@type”:”Question”,”name”:”Should a 22-year-old avoid all roles with low AI Stability?”,”acceptedAnswer”:{“@type”:”Answer”,”text”:”Not necessarily. Some low-score roles are still on-ramps to higher-durability senior roles. The risk is treating a low-score role as a 30-year path.”}}
]}
]}
Last updated:
Written by The DreamJobMatcher Editorial Team, Career Assessment Specialists with backgrounds in occupational psychology and HR · About the team
Related: AI career test: find a job AI will not replace →
Figures attributed to DreamJobMatcher, including AI Stability Scores and fit-score percentages, are estimates produced by our internal framework and models. They are shown for general guidance and illustration. They are not outcomes from a controlled or peer-reviewed study, and individual results vary.
