Facial attractiveness and profile-based perceived orthognathic surgery need: comparative assessment among professionals, laypersons, and artificial intelligence models.

Abstract

This study evaluated the effect of controlled 2° sagittal profile modifications on facial attractiveness ratings and profile-based perceived orthognathic surgery need using one standardized female and one standardized male facial model. For each model, nine digitally manipulated profile images were generated from -8° to +8° relative to baseline and assessed by orthodontists, oral and maxillofacial surgeons (OMFS), and laypersons (n = 72). Two multimodal artificial intelligence systems, ChatGPT-4.0 and Gemini 1.5 Pro, independently generated attractiveness scores to provide an exploratory benchmark for human-AI agreement. The selected models showed distinct sagittal response patterns. For the female model, attractiveness ratings peaked within a mild retrusive range (-2° to -4°), whereas ratings for the male model increased with greater sagittal projection. Significant intergroup differences were observed at selected sagittal positions, with orthodontists generally showing narrower esthetic tolerance. Profile-based surgical recommendations differed significantly for the male model (P < 0.001), with OMFS specialists recommending intervention most frequently, while no significant overall group difference was found for the female model. Human-AI agreement was model-dependent: Gemini showed closer alignment with human ratings for the female model, whereas ChatGPT-4.0 aligned more closely with human ratings for the male model. Both AI systems declined to provide surgical recommendations because of built-in ethical restrictions. Within the limitations of a two-model, profile-based design, sagittal esthetic perception appeared to vary according to model sex, evaluator background, and AI system. These findings should be interpreted as model-specific and exploratory rather than as generalizable sex-based esthetic norms.