Every career on this site carries an AI resilience score from 0 to 100. Other pages — the Best Overall ranking, individual career profiles, the career quiz — use that score as one of six weighted factors. This page explains what the score actually measures, where the data comes from, and why high AI exposure does not mean "about to be replaced."
The underlying methodology comes from Pew Research Center’s 2023 analysis of AI exposure across occupations. Pew classified 41 work activities from the O*NET Database as high, medium, or low exposure to AI capabilities. We adapted their cross-industry framework for intra-industry comparison across the published health care careers in our dataset, adding health care-specific overlays for state licensure, radiology, and medical coding.
The critical distinction: exposure does not equal replacement. Nurse practitioners have high AI exposure — they spend significant time on documentation, clinical decision-making, and data interpretation, all areas where AI tools are advancing rapidly. But NPs are using AI ambient scribes, clinical decision support, and automated prior authorization to see 20–30% more patients. Their AI exposure is making them more valuable, not less. Meanwhile, CNAs have very low AI exposure because their work is physical patient care in uncontrolled environments. They are not "safer from AI" — they are simply doing different work.
Three dimensions shape how AI actually affects a health care career: task displacement (can AI do part of this job cheaper?), role amplification (does AI make this worker more productive?), and regulatory protection (do licensing requirements prevent AI from operating independently, even where it’s technically capable?). Our AI resilience score captures all three. The editorial context on individual career pages goes deeper.
How the Score Works
What you do every day
The federal O*NET Database breaks every career into 41 work activities — things like "documenting information," "performing physical activities," and "making decisions." Research from Pew Research Center classifies each activity as high, medium, or low overlap with what AI can currently do. Careers built around physical patient care — repositioning, wound care, manual therapy — score high on resilience because those tasks have very low AI overlap. Careers built around data processing and documentation score lower.
Whether AI has already arrived
Some health care fields already have production-scale AI. Radiology imaging AI reads scans at specialist-level accuracy. Medical coding algorithms process claims faster than human coders. When AI is not theoretical but already deployed in a field, the resilience score accounts for that reality with a downward adjustment. Careers where AI remains experimental or absent do not receive this penalty.
What the law says
State-licensed careers have a regulatory barrier that pure task analysis misses. Even where AI can technically perform the work, it cannot hold a license. A nurse practitioner’s scope of practice is defined by state law, and those laws change slowly regardless of what the technology can do. Licensed careers receive a resilience boost because this legal protection is real and durable — not because the underlying work is less exposed to AI.
Key Findings
- •Physical patient contact correlates with lower exposure: The highest-scoring careers in the current dataset cluster in hands-on clinical and emergency-response work. O*NET activity patterns consistently place these tasks in lower-exposure categories.
- •State licensure creates regulatory barriers: All top-scoring careers require state licensure or certification, creating regulatory barriers that prevent AI from operating independently — even when the technology exists. Scope-of-practice laws move slowly regardless of technical capability.
- •Administrative roles show higher exposure and weaker outlooks: Medical transcription and coding-heavy roles sit near the lower end of resilience scoring and also align with BLS roles facing decline or slower hiring. The overlap is about task structure, not title alone.
- •AI amplification still matters: Nurse practitioners, physician assistants, and surgeons still carry meaningful AI exposure in documentation and decision-support tasks, but adoption can raise productivity rather than replace the role.
AI Risk Levels for All Health Care Careers
All published careers classified by AI risk level, from Very Low AI Risk (most resilient) to Very High AI Risk (most exposed). Risk levels combine task overlap, regulatory protection, and demonstrated AI capability in the field. Careers marked with * use task-derived scoring (see methodology).
| Rank ↑ | Career ↕ | AI Risk Level ↕ | Growth % ↕ | Salary ↕ | Education ↕ | |
|---|---|---|---|---|---|---|
#1 | Emergency Medical Technician (EMT) | Very Low AI Risk | 5.1% Average | $39k/yr | Postsecondary Certificate | View Details |
#2 | Paramedic | Very Low AI Risk | 5.1% Average | $39k/yr | Associate Degree | View Details |
#3 | Physical Therapist Assistant | Low AI Risk | 14.2% Fast | $64k/yr | Associate Degree | View Details |
#4 | Dental Hygienist | Low AI Risk | 6.9% Average | $94k/yr | Associate Degree | View Details |
#5 | Registered Nurse | Low AI Risk | 6.2% Average | $94k/yr | Associate Degree | View Details |
#6 | Occupational Therapy Assistant | Low AI Risk | 9.9% Fast | $67k/yr | Associate Degree | View Details |
#7 | Certified Registered Nurse Anesthetist (CRNA) | Low AI Risk | 9.0% Fast | $223k/yr | Doctoral Degree | View Details |
#8 | Certified Nursing Assistant (CNA) | Low AI Risk | 4.3% Average | $38k/yr | Postsecondary Certificate | View Details |
#9 | Radiation Therapist | Low AI Risk | 1.2% Below Average | $98k/yr | Associate Degree | View Details |
#10 | Certified Nurse-Midwife (CNM) | Low AI Risk | 11.0% Fast | $129k/yr | Master's Degree | View Details |
#11 | Occupational Therapist | Low AI Risk | 10.5% Fast | $97k/yr | Master's Degree | View Details |
#12 | Nuclear Medicine Technologist | Low AI Risk | 3.2% Average | $93k/yr | Associate Degree | View Details |
#13 | Respiratory Therapist | Low AI Risk | 8.4% Fast | $78k/yr | Associate Degree | View Details |
#14 | Audiologist | Low AI Risk | 10.0% Fast | $92k/yr | Doctoral Degree | View Details |
#15 | Licensed Practical Nurse (LPN) | Low AI Risk | 4.7% Average | $60k/yr | Certificate or Diploma | View Details |
#16 | Orthotist and Prosthetist | Low AI Risk | 9.8% Fast | $78k/yr | Master's Degree | View Details |
#17 | Athletic Trainer | Low AI Risk | 17.3% Very Fast | $60k/yr | Master's Degree | View Details |
#18 | Physical Therapist | Low AI Risk | 14.2% Fast | $101k/yr | Doctoral Degree | View Details |
#19 | Emergency Medicine Physician | Moderate AI Risk | 3.0% Average | $321k/yr | Doctoral Degree | View Details |
#20 | Nurse Practitioner | Moderate AI Risk | 40.0% Very Fast | $129k/yr | Master's Degree | View Details |
Sources
- •O*NET Database 28.0 — Work Activities and Work Context
- •Pew Research Center: Which U.S. Workers Are More Exposed to AI on Their Jobs, 2023
- •Bureau of Labor Statistics — Occupational Outlook Handbook, 2024–34
- •Bureau of Labor Statistics — OEWS, May 2024
- •McKinsey & Company: Transforming Healthcare with AI, 2020
- •American Medical Association: How AI Is Helping Doctors Improve Care
Methodology
Data source. AI resilience scores are calculated using O*NET Database 28.0 work activity profiles. Each occupation in ONET has importance and level ratings for 41 Generalized Work Activities (GWAs). We mapped each published health care career in our dataset to its corresponding ONET-SOC code and extracted the full work activity profile.
Pew Research tier classification. We adopted the Pew Research Center’s AI exposure framework which classifies the 41 GWAs into three tiers based on overlap with demonstrated AI capabilities: 16 HIGH exposure activities (e.g., analyzing data, processing information, documenting information), 16 MEDIUM exposure activities (e.g., making decisions, interpreting meaning, scheduling), and 9 LOW exposure activities (e.g., performing physical activities, handling objects, operating vehicles).
Scoring formula. For each career, we calculate a weighted composite: `score = (low_contribution × 1.5) + (medium_contribution × 0.5) - (high_penalty × 2.0) + 50`. Low-exposure activities contribute positively (physical, hands-on work). High-exposure activities penalize (data processing, documentation). The offset normalizes the scale to roughly 0–100, though observed values in the current published set are concentrated in a narrower band.
Health care overlays. Three domain-specific adjustments account for factors the generic Pew methodology does not capture: state-licensed roles receive +15 points (regulatory barriers prevent AI from operating independently regardless of technical capability); radiology-focused roles receive -20 points (demonstrated AI superiority in pattern recognition for imaging); medical coding roles receive -30 points (rule-based classification AI already performs at production scale).
Task-derived scoring. Six careers (Dialysis Technician, General Surgeon, Sleep Technologist, Sterile Processing Technician, Patient Care Technician, Medical Laboratory Assistant) lacked direct ONET-SOC mappings. For these, we used LLM-assisted classification of their primary work tasks against the 41 GWAs, then applied the standard scoring formula. These careers are marked with an asterisk () in the data table.
Limitations. This methodology measures task-level overlap with AI capabilities, not prediction of job displacement. A career with high AI exposure may benefit enormously from AI adoption (as with NPs and ambient scribes) or face genuine displacement risk (as with medical transcriptionists). The Pew framework was designed for cross-industry comparison; our adaptation for intra-industry health care comparison adds precision but cannot account for pace of adoption, institutional inertia, or workforce supply dynamics. Scores should be read as one input among several, not as standalone career guidance.