The study, led by Omeed Modiri of the David Geffen School of Medicine at UCLA together with colleagues at the Icahn School of Medicine at Mount Sinai, including senior author Jesse M. Lewin, surveyed members of the American College of Mohs Surgery, the professional home of physicians who perform Mohs micrographic surgery. This technique, considered the gold standard for many skin cancers, involves removing thin layers of tissue and examining each one under a microscope until no cancer cells remain. It is a discipline that demands precision, real-time pathological judgment, and careful reconstructive planning, making it an especially revealing setting in which to gauge how surgical specialists perceive AI.

The researchers distributed an anonymous twelve-question survey to ACMS members and analyzed the responses using Fisher’s exact tests, with a p value below 0.05 considered statistically significant. Thirty-six surveys were completed, a modest sample that the authors themselves flag as a key limitation alongside sampling and recall bias inherent to survey-based research. Even so, the results sketch a coherent picture of a specialty in transition. Forty-two percent of respondents reported having already used AI tools in some capacity, and sixty-seven percent of all participants said they were likely to incorporate AI into their clinical practice within two years. In other words, adoption is no longer a hypothetical question for Mohs surgeons; it is a matter of when and how.

The platforms these surgeons gravitate toward say a great deal about what they currently trust AI to do. ChatGPT dominated, used by ninety-six percent of those who reported AI experience, followed by OpenEvidence at thirty percent and Gemini at twenty-two percent. Custom AI models, the kind purpose-built for specific clinical tasks, saw only limited use. This pattern suggests that Mohs surgeons are relying on general-purpose large language models and medical literature search tools rather than bespoke surgical applications, treating AI today less as a scalpel and more as an assistant for information gathering and workflow support.

When asked about the benefits, respondents pointed overwhelmingly to efficiency rather than clinical prowess. Seventy-six percent identified increased efficiency as the most significant advantage of AI in dermatologic surgery, while sixty-two percent cited reduced administrative burden. These figures align with a broader theme in medical AI research: the technology’s most immediate and reliable value often lies in taming the paperwork, documentation, and coding tasks that consume clinician time, rather than in making high-stakes decisions. Prior work by some of the same authors, for example, evaluated how accurately ChatGPT and Gemini assign Current Procedural Terminology billing codes in dermatologic surgery, a quintessentially administrative application where errors are recoverable and oversight is straightforward.

The concerns, however, were sharper and more consistent than the optimism. Eighty-six percent of respondents worried about inaccurate or unreliable information, the highest of any concern measured in the survey. Fifty-four percent flagged patient safety and liability, and forty-nine percent feared the dehumanization of care. These numbers reveal a profession that understands exactly where the fault lines lie. Large language models can generate fluent, confident text that is nonetheless wrong, a phenomenon that becomes far more dangerous when the output informs where to cut tissue or how to reconstruct a face. Liability questions compound the technical ones, since it remains legally murky who bears responsibility when an AI-assisted recommendation contributes to a poor surgical outcome.

Nowhere is this tension more visible than in the gap between expectation and comfort. Seventy-two percent of participants believed AI will change the practice of dermatologic surgery over the next five to ten years, yet only thirty-nine percent reported feeling any degree of comfort with AI-assisted tools guiding surgical margins or repairs. The surgeons, in effect, are saying that transformation is coming whether they like it or not, while drawing a firm line at delegating intraoperative judgment. When it came to patient outcomes, opinions split further: fifty percent believed AI would lead to improvement, nineteen percent expected a neutral effect, twenty-five percent were unsure, and six percent anticipated harm. That distribution captures a specialty that is cautiously hopeful but far from convinced.

Experience appears to shape attitudes, though the survey could not prove it statistically. Surgeons with fewer than five years in practice were 2.83 times more likely to report having used AI tools compared with those in practice longer, a substantial numerical difference that fell short of statistical significance with a p value of 0.21, likely a casualty of the small sample size. Still, the trend fits a familiar pattern across medicine: younger clinicians, trained alongside consumer AI tools, arrive in practice already fluent in them, while more senior surgeons encounter the technology later and often more skeptically. If the trend holds in larger samples, the demographic gradient of AI adoption could become a defining feature of how surgical specialties modernize.

The study’s technical limitations deserve honest weight. Thirty-six respondents represent a small fraction of the ACMS membership, and those motivated to answer a survey about AI are plausibly more interested in the topic than their peers, introducing sampling bias. Recall bias may also color self-reported usage patterns. The authors are transparent about these constraints and call for larger studies to validate the findings, as well as for standardized guidelines evaluating responsible AI use in healthcare. Their caution echoes wider calls in the literature, including proposals for formal requirements to mitigate patient harm risks from AI in medicine and frameworks for adopting large language models safely in clinical settings.

What emerges from this report is not a story of resistance or revolution but of calibrated pragmatism. Mohs surgeons, whose work demands microscopic precision and carries visible, personal consequences for patients, are welcoming AI into the margins of their practice, the scheduling, the documentation, the literature searches, while keeping it at arm’s length from the margin of excision. That distinction may prove to be the template for surgical AI adoption more broadly: efficiency gains first, decision support last, and rigorous validation, clear liability frameworks, and preserved human judgment as the non-negotiable conditions in between. As one of the first systematic looks at AI attitudes within this specialty, the survey suggests that the next five to ten years will test whether AI developers can earn the trust that these surgeons are not yet ready to give.

Subject of Research: A survey of American College of Mohs Surgery members on the current use, perceived benefits, and risks of artificial intelligence in dermatologic surgery.

Article Title: Current impact and future outlook of artificial intelligence among American College of Mohs Surgery members

Article References: Modiri, O., McGeough, O., Elder, A., Ensslin, C., & Lewin, J. M. (2026). Current impact and future outlook of artificial intelligence among American College of Mohs Surgery members. Archives of Dermatological Research, 318(1), Article 458. https://doi.org/10.1007/s00403-026-04791-5

Image Credits: AI Generated

DOI: 10.1007/s00403-026-04791-5

Keywords: artificial intelligence, Mohs surgery, dermatologic surgery, ChatGPT, large language models, survey, skin cancer, clinical practice, patient safety, surgical margins, healthcare workflow, American College of Mohs Surgery