Is Generative Artificial Intelligence (AI) Capable of Proposing Adequate Treatment Plans for Dentofacial Deformities?

Background: The use of artificial intelligence remains limited in the field of Craniomaxillofacial Surgery. The purpose of this article is to determine whether public-access AI has the potential to accurately diagnose and propose treatment plans in patients with dentoskeletal deformities undergoing orthognathic surgery. Methods: This was a multicenter retrospective cohort study consisting of patients (n = 150) with a dentoskeletal deformity, divided into classes I, II, and III. NemoFab and Dolphin were used for data collection, and three queries were posed to ChatGPT (model version 4.0): Question 1 (Q1) was for diagnosis; Question 2 (Q2) was for surgery-type identification; and Question 3 (Q3) was for computer-assisted surgical planning. A Kruskal–Wallis test was used to analyze the statistical differences between queries and Kendall’s Tau-B test was used to identify associations. A p-value < 0.05 (95%CI) was considered statistically significant. Results: Q1 had an accuracy of 84.65%, Q2 was 90.5%, and Q3 was 63.59% accurate. The Kruskal–Wallis test was statistically significant (p < 0.0001). Kendall’s Tau-B showed statistically significant associations between Q1 and Q2 (p < 0.0001) and between Q1 and Q3 (p < 0.006), but not between Q2 and Q3 (p = 0.545). Conclusions: Public-access AI currently has the ability to recognize diagnostic patterns and identify previously performed surgical procedures; however, it remains insufficiently reliable for virtual surgical planning.