Why bone work stays with people
Orthopedic surgery is physical problem-solving inside a living body. A surgeon reduces a fracture, cuts bone to a plan, seats an implant, then judges soft-tissue tension by feel before closing. None of that is a text problem. It is force, tissue quality and timing, decided in the moment with a team watching.
The question “will AI replace orthopedic surgeons” usually gets asked because robotic arms are already common in knee and hip replacement. Those systems hold a cutting guide steady inside limits a surgeon sets. The plan, the exposure, the implant choice and every change of plan mid-case still come from the person scrubbed in.
Before the operating room, there is more human work. Examining a painful joint, weighing surgery against physical therapy, explaining risk to a patient who is frightened, and taking consent are judgment and relationship tasks. Our coverage score, which estimates the share of task time AI can handle today, sits at 18 out of 100 for this job. You can read how that figure is built on the coverage method page.
What AI handles, what it assists, what it leaves alone
The tasks AI can take on by itself are the paperwork-shaped ones around the case: drafting operative notes and clinic letters, pulling prior imaging and records into order, and flagging findings on X-rays and MRIs for a human to confirm. That group accounts for 0% of task time on this page.
A larger band of work is assisted rather than handed over. Preoperative planning software suggests implant sizing and alignment targets; a robotic platform keeps a saw inside a planned boundary; risk models rank which patients may struggle after discharge. The surgeon approves, overrides or ignores each one. Assisted work covers 40% of task time.
What is left needs a trained person in the room: the operative steps themselves, revision surgery when anatomy does not match the scan, trauma cases that arrive without a plan, and the conversations that decide whether to operate at all. That group is 60% of task time, and it is the reason the headline score lands where it does.
How strong is the evidence?
Thin, and we grade it honestly. The quality-parity evidence grade for this job is D, which means no study has tested an AI system against a qualified orthopedic surgeon on the operative work itself. So we publish no parity number here. Research on image reading and outcome prediction exists, but reading a scan is not performing an arthroplasty.
What would settle it: prospective, multi-center trials comparing an autonomous system with surgeons on the measures the field already tracks, such as component alignment, infection and complication rates, revision at two and five years, and patient-reported function. Until results like those are published and repeated, a number would be guesswork. The quality-parity method page explains what each grade from A to D requires.
When the picture could shift
Most likely after 2042 (8 in 10 of our scenarios). We do not restate what that window measures here; the replacement-year method page sets out how the range is produced and why we publish a span rather than a date.
Two things could pull it earlier. First, hardware: the robotics panel above puts a large share of this job in the physical column, and the class of machine needed is at the dexterous-humanoid end, so progress in manipulation hardware matters more here than progress in language models. Second, cost. Assisting software is cheap next to a surgeon’s time; the cost panel on this page shows that gap clearly, and cheap tools spread fast.
Two things hold it back. Liability and regulation are the obvious ones: approval for a system that operates without a surgeon supervising is a different bar from approval for a guided cutting tool. Case variability is the quieter one. Trauma, deformity, failed implants and poor bone stock throw up situations no plan covered, and recovering from those is the skill the job is paid for.
Good to know: the US Bureau of Labor Statistics projects employment for this occupation to grow about 4% between 2025 and 2035, with around 14,100 people employed and median pay near $358,550 (BLS, 2025).
How to stay needed in this job
Lean into the work that stays human. Take on revision and complex trauma cases, where the plan changes under your hands. Own the decision to operate or not, including the cases you turn down. Keep the consent conversation yours, in plain words, with the patient’s goals in it.
Two skills pay off alongside that. One is reading machine output critically: knowing when a planning suggestion or a flagged scan finding is wrong, and being able to say why. The other is teaching, since residents still learn hands and judgment from people, and supervising that transfer is hard to hand off.
Nearby work is scored too. Compare this page with pediatric surgeons, sports medicine physicians and radiologists, whose task mix leans much harder on image interpretation. The wider diagnosing and treating practitioners family and the healthcare sector page show how those scores cluster.
This job’s Still needs a human score is 78 out of 100 (higher is safer). To see how it was built, read how the scoring works. To see where it sits among jobs that mostly need a person (our top band, Nah.) and the bands below it, open the safest jobs list, or put two jobs side by side with the comparison tool.