Review finds no clear overall advantage for AI surgical training

A systematic review and meta-analysis of randomized controlled trials found no statistically significant overall advantage for artificial intelligence-assisted training in improving short-term surgical technical skills. Across nine studies enrolling 588 participants, four studies contributing data from 154 participants entered the primary meta-analysis. The pooled effect was 0.40 in Hedges’ g, with a 95% confidence interval from −0.30 to 1.10 and P = 0.17. The result does not establish that AI-assisted training improves operative technical performance overall.

The review, published in BMC Medical Education, evaluated randomized studies of AI-assisted training for acquiring operative technical skills. AI-assisted systems in this literature were assessed for automated evaluation of performance and individualized feedback. The researchers searched PubMed, Embase.com, the Web of Science Core Collection, the Cochrane Central Register of Controlled Trials and IEEE Xplore from their inception through 6 August 2026. They also conducted backward and forward citation searches on 7 August 2026. Overall technical skill performance was the primary outcome, and effect estimates were expressed as Hedges’ g and pooled using a random-effects model.

Results differed according to the comparison condition. Against active human instruction, AI-assisted training had a pooled standardized mean difference of 0.13, with a 95% confidence interval from −0.24 to 0.51 and P = 0.48. This comparison therefore did not demonstrate superiority over instruction provided by human experts. When the AI-assisted programs were evaluated against passive conditions or training without AI, the pooled effect was larger: 0.94, with a 95% confidence interval of 0.30 to 1.59 and P = 0.004. The difference between the two subgroup estimates was statistically significant at P = 0.03, although each subgroup contained only two studies. The authors consequently described the larger estimate against passive conditions or training without AI as exploratory rather than definitive.

The evidence was also limited by methodological concerns. Statistical heterogeneity in the primary analysis was 36%, as measured by I². Sensitivity analyses did not materially change the interpretation of the results. Risk of bias was assessed with the Cochrane Risk of Bias 2 tool: one study was rated at low overall risk of bias, while eight had some concerns. Using the Grading of Recommendations Assessment, Development and Evaluation approach, the researchers rated the certainty of the evidence as low because of risk of bias and imprecision. These limitations mean that the numerical effect estimates should not be treated as precise measurements of AI’s educational value.

The findings do not show that AI-assisted surgical training is ineffective, nor do they establish that it improves long-term clinical outcomes. The primary endpoint was short-term technical skill performance, and the included evidence did not demonstrate an overall benefit. Nor did the review show that AI-assisted instruction could replace active expert teaching, because the comparison with human instructors produced no statistically significant advantage. At the same time, the larger point estimate against passive conditions or training without AI leaves open the possibility that automated assessment and individualized feedback may offer benefits under some circumstances. The authors concluded that the current evidence is compatible with a complementary, rather than replacement, role for AI in surgical education and called for larger trials with standardized outcomes and longer follow-up.

Subject of Research: Randomized evidence on AI-assisted training for acquiring operative technical skills in surgery

Article Title: Artificial intelligence-assisted surgical skill acquisition: a systematic review and meta-analysis of randomized controlled trials

Article References: Cai, M., Luo, B., Xiang, Q., & Wang, P. (2026). Artificial intelligence-assisted surgical skill acquisition: a systematic review and meta-analysis of randomized controlled trials. BMC Medical Education. https://doi.org/10.1186/s12909-026-10257-z

Image Credits: AI Generated

DOI: 10.1186/s12909-026-10257-z

Keywords: artificial intelligence, surgical education, technical skills, randomized trials, meta-analysis, expert instruction, automated feedback, evidence certainty, operative training, risk of bias, Artificial, intelligence-assisted