Free sampler. Re-scored every six months.
Medicine is among the best-protected careers we track, but the score depends heavily on specialty. Surgery scores 2.1 out of 10 for AI exposure — the lowest of any career in this index. Primary care scores 2.9. Radiology scores 4.9, and it is the fastest-moving specialty we measure. The protection in medicine is physical presence and human accountability, and specialties that rely on neither are moving.
The short answer for parents: yes, medicine remains one of the safest paths available — but for the first time, choice of specialty carries an AI dimension. A student entering medical school today will choose a specialty in five or six years, and the specialties are no longer moving together.
Every career in this index is scored 1–10, where 10 is most exposed to AI. Same six factors, same weights, applied identically to a surgeon and a paralegal.
| Medicine track | 2023 | 2025 | Now | 3-yr move | Band |
|---|---|---|---|---|---|
| Surgery | 1.8 | 1.9 | 2.1 | +0.3 | Very low |
| Emergency medicine | 2.0 | 2.2 | 2.5 | +0.5 | Very low |
| Primary care / family medicine | 2.4 | 2.6 | 2.9 | +0.5 | Low |
| Anesthesiology | 2.7 | 2.9 | 3.1 | +0.4 | Low |
| Psychiatry | 3.1 | 3.4 | 3.7 | +0.6 | Low |
| Pathology | 3.8 | 4.3 | 4.7 | +0.9 | Moderate |
| Radiology | 3.9 | 4.4 | 4.9 | +1.0 | Moderate |
2023 and 2025 figures are reconstructed using current methodology, not archived from past editions.
For scalethe median career in this edition scores around 5.5. Bedside nursing scores 2.8. Physical therapy scores 2.5. Entry-level software development scores 8.1.
Surgery at 2.1 is the lowest score in the entire index — lower than nursing, the trades, or any other career we measure.
The spread is the story. Three years ago the gap between the safest and most exposed medical specialty was 2.1 points. It is now 2.8, and every year it widens. Medicine is not becoming riskier. It is becoming less uniform.
No. But it is already changing what several specialties do all day.
The pattern is consistent across the profession: AI is absorbing the parts of medicine that involve reading, matching and documenting, and leaving the parts that involve a body in a room and a human being accountable.
For most doctors this is welcome. Documentation burden is among the most-cited causes of physician burnout, and ambient systems that draft the note from the consultation are already in routine use.
The specialties where it cuts deeper are the ones whose core work is reading and matching.
How primary care rates against each. Ratings are 0–10 on each factor's own terms.
Medicine holds the strongest protection stack of any career we score. What varies between specialties is the first factor.
So you can see what the analysis actually looks like.
How much of this job can AI already do? — rated 6.0
How much of this job can AI already do? — the factor that separates the specialties
This factor carries 35% of the score, more than any other, and it is where medicine's internal spread comes from.
Primary care rates 6.0. A substantial share of the work is documentation, coding, correspondence and guideline application — all of it automating quickly. What resists is the physical examination, the relationship built over years, and the judgment call on a patient who presents atypically. The automatable share is large but it is the administrative shell rather than the clinical core.
Surgery rates 4.5. Planning, imaging review and documentation are automatable. The operation is not, and nothing on the visible horizon changes that. This is why surgery sits at the bottom of the entire index.
Radiology rates 8.3, up from 5.6 in 2023. Image interpretation is pattern recognition on structured visual data, which is the single thing AI does best. Around three-quarters of all FDA-authorized AI medical devices are in radiology — the technology is concentrated there because that is where it works. The radiologist's job is shifting from reading to supervising, adjudicating complex cases, and owning the decision.
That is a 3.8-point spread on one factor, inside one profession, held by people with the same medical degree. It is the clearest example in this index of why scoring professions as single units hides the thing that matters.
Ranked by exposure, safest first:
The pattern holds across every profession we score: move the same work behind a screen and exposure roughly doubles. Nursing goes from 2.8 at a bedside to 5.4 at a desk. Teaching goes from 3.6 in a classroom to 6.8 online.
This is the question we get asked most, and our honest answer is that we are less confident about this score than any other in the index.
The case for concern is real. Image interpretation is exactly what AI does well, adoption is furthest advanced here, and the score has moved a full point in three years — faster than any other protected career we track.
But the labor market has done the opposite of what the narrative predicts. Radiology posts sit unfilled, residency programs match at close to 100%, imaging volumes keep rising, and the shortage has deepened over the same decade in which displacement was being confidently forecast.
Both things are true, and the gap between them is wider in radiology than anywhere else we score. We publish the number with the disagreement stated rather than adjusting it to match the employment data — because bending a score to fit an outcome would make every trend in this index meaningless.
The full profile sets out both cases properly, including the strongest published argument against our position.
The AI question and the career question are different, and this index only answers the first.
Medicine scores exceptionally well on exposure, and its entry route — residency — is the most protected on-ramp we measure, mandated and structurally impossible to automate away.
The reasons people hesitate about medicine have not changed and have nothing to do with AI: it is a decade or more to independent practice, the training is expensive, early-career hours are punishing, and most systems require relocating repeatedly for training posts. Burnout affects around 42% of physicians — high, though it has fallen three years running, from 48.2% in 2023. A fifteen-year-old should understand that timeline concretely before committing to it.
A low AI score does not mean an easy path. The full profile covers the downsides honestly, alongside what doctors themselves say about the work.
This sampler tells you where medicine stands. The full profile tells you what to do about it.
Most families are weighing two or three careers seriously, and a few more they haven’t ruled out. Pick the ones you need.
Each includes a short version written directly to the student and the technical scoring appendix. Spring 2027 updates of whatever you buy are included.
Read one complete profile free → We publish computer science in full so you can judge the depth before buying anything.