How Do Surgeons Feel About AI in the Operating Room?
Updated: Sep 20
A survey from University Medical Center Utrecht puts a number on something surgeons have said quietly for years.
Fifty-three surgeons across five continents and nineteen countries rated five types of intraoperative AI assistance.
Anatomy recognition and risk detection scored highest. Step recognition and decision-making guidance scored lowest.
The message underneath the numbers: surgeons want AI to see clearly. They don't want it to decide.
The Gap Between Belief and Use
86.8% of respondents agreed intraoperative AI could improve surgical performance.
75.5% said they'd trust a clinically validated AI tool.
79.2% said they had never used one during surgery.
The belief is there. The access isn't.
What Surgeons Actually Want
Participants watched paired surgical footage from robot-assisted esophagectomy recordings, one unaltered and one with an AI overlay.
Anatomy recognition scored 4.57 out of 5. Risk detection followed at 4.45.
Vision-language model guidance came next at 3.94. Decision-making assistance scored 3.51. Step recognition scored lowest, at 3.36.
The differences were statistically significant. Surgeons don't reject AI uniformly. They rank it by what it touches.
Experience Didn't Change the Answer
Novices, intermediates, and experts rated the tools within a narrow band of each other.
No significant difference showed up across experience levels for knowledge, confidence, or perceived usefulness.
The one place experience mattered: 90.6% of all respondents said AI would help trainees and fellows most. Only 69.8% said the same about consultants.
Where This Leaves the Industry
The paper's own conclusion is blunt. Adoption is limited by availability, not appetite.
Surgeons are asking for tools that watch the field and flag risk. Not tools that choose the next step.
That's a narrower brief than most surgical AI roadmaps are currently building toward.
