Why deliveries and surgery stay with people
Obstetrics and gynecology mixes two kinds of work that machines handle poorly. One is physical: catching a baby, repairing a tear, running a cesarean delivery, operating in a pelvis that does not match the textbook. The other is judgment under time pressure, with a patient and a family in the room and minutes to decide.
The task list on this page splits the job along those lines. Documentation, coding and the first pass over images and lab results are the parts software already touches. Intrapartum decisions, operative work and face-to-face counseling about contraception, fertility or a loss are the parts that stay with a clinician who is legally and personally accountable for the outcome.
Hardware is the other limit. Our robotics estimate places the physical side of this job in the dexterous humanoid tier, which means a machine would need hands and reflexes good enough for an unplanned delivery, not just a steady arm on a rail. Surgical robots today are controlled by a surgeon; they extend reach rather than remove the operator. So when people ask will AI replace obstetricians, the honest answer starts with task erosion around the edges, not an empty labor ward.
What AI does, what it helps with, and what stays human
Start with the share AI can take on its own: 9% of task time sits in the group where a tool can finish the job. That is mostly paperwork-shaped work — drafting clinic notes from a recorded visit, assembling discharge summaries, pulling structured data out of records and pushing it into billing codes. None of it touches a patient.
Next, the assist group: 33% of task time is work where software speeds a physician up but does not sign off. Automated fetal biometry measures on an ultrasound image, flagging an abnormal trace, surfacing prior results during a prenatal visit, checking a plan against guidelines. The clinician still reads, confirms and decides. Our coverage score — the share of task time AI can handle today — comes from this split and is explained on the coverage method page.
Then the part that keeps needing a person: 58% of task time. Performing operative deliveries and gynecologic surgery sits here. So does managing a labor that turns complicated, examining a patient, and talking through options when the evidence is balanced and the choice is hers. Those tasks are the reason the headline score lands where it does.
What the evidence actually tests
There is no head-to-head test of AI against obstetricians on this job’s real work in our evidence set, which is why the quality-parity grade is D. A grade at that level means not measured, so this page gives no parity number. Image-reading and note-drafting studies exist across medicine, but reading a scan is one task inside a job built from many.
What would settle it is specific: prospective studies in obstetric and gynecologic care that compare AI-led interpretation — fetal growth measurement, cardiotocography review, cervical screening triage — against board-certified physicians, with patient outcomes as the endpoint rather than agreement with a label. Add measured effects on clinician time and error rates in live clinics. Until work like that is published and repeated, the honest position is uncertainty, graded as such. Our full approach is set out in the scoring methodology.
When the picture could change
Most likely between 2041 and 2060 (8 in 10 of our scenarios). What that window measures, and how it is built, is described on the replacement-year method page.
Two things could pull it earlier. Ambient documentation tools are spreading fast through hospital systems, and once a tool is trusted for notes it tends to creep toward triage and order drafting. Automated measurement on ultrasound is also maturing, which could move a slice of scan review out of physician hands and into a technologist-plus-software workflow.
Two things hold it back. First, accountability: nobody has solved who answers for an intrapartum decision made by software, and consent and liability rules move slowly. Second, the physical work. Autonomous surgery and autonomous delivery need hardware that does not exist outside research labs, and hospital procurement, credentialing and training add years on top of any technical win. Demand matters too: BLS projects employment for obstetricians and gynecologists to change by about 1.7% between 2025 and 2035, from roughly 21,260 jobs, with median pay of $292,910 (BLS, 2025). That is a flat-to-slightly-growing field, not a shrinking one.
How OB-GYNs stay needed
Lean into the tasks in the needs-a-person group. Operative and intrapartum care is the clearest one: surgical volume and complication management are hard to delegate to a tool. Complex counseling is the second — fertility, contraception, pregnancy loss, surgical options — where the work is helping someone choose, not producing an answer. Supervising and teaching is the third: residents, midwives and sonographers all need a clinician who can judge both the case and the software’s output.
Two skills compound. One is verification: knowing how a model fails on an image or a draft note, and building a habit of checking it. The other is leading mixed teams and protocols, so you shape how these tools enter your unit instead of inheriting someone else’s setup.
What to do: ask who reviews, signs and takes responsibility for every AI output in your clinic, and get that written into the workflow.
Nearby work follows the same pattern. Compare this page with Family Medicine Physicians, Pediatricians, General and Nurse Midwives, where documentation is the first thing to move and hands-on care is the last. You can also see the wider group on the healthcare diagnosing and treating practitioners page, the industry view in healthcare, and how this job sits among the jobs least exposed to AI. To line it up against one other role, use the side-by-side comparison tool.