Comparative evaluation of artificial intelligence-assisted smile designs: Aesthetic preferences among dental...

To evaluate aesthetic preferences for artificial intelligence (AI)-assisted smile designs produced using four digital smile design software platforms, and to explore associations between these preferences and demographic factors, among dental specialists and laypeople.

A cross-sectional study was conducted to assess aesthetic preferences for AI-assisted smile designs among dental specialists and laypeople. Photographs of 10 standardized models were processed through four digital smile design software platforms: Smilefy, Smilecloud, DTS PRO, and Trios Smile Design. A single prosthodontist reviewed each automated output and adjusted tooth shape, color, midline, gingival contour, and occlusal plane where needed to conform to clinical norms; outputs are therefore best described as AI-assisted rather than fully automated. Each output was labeled A-D to blind participants to the software used; this letter assignment was shuffled independently for each of the 10 model sets rather than following a single fixed layout, though the specific mapping used per set was not formally logged. Participants ranked the four outputs per model from 1 (most aesthetic) to 4 (least aesthetic) using a blinded, pretested questionnaire. Of the participants recruited, at least partial ranking data were available for an analytic sample of 78 dental specialists and 86 laypeople (164 total; demographic data were available for 76 and 85, respectively). Because rankings were ordinal, clustered within participants, and forced-choice, aesthetic preference was analyzed using the Friedman test with Kendall's coefficient of concordance (W) as an effect size, followed by Wilcoxon signed-rank pairwise comparisons with Holm correction. Associations between preference and professional group, age, and gender were examined using Mann-Whitney U and Kruskal-Wallis tests (Holm-corrected), applied to each software's mean rank.

Rankings differed significantly across the four software platforms (Friedman χ² = 349.8, df = 3, p < 0.001; Kendall's W = 0.71, indicating strong agreement). Smilefy and Smilecloud were the most preferred designs overall (mean ranks 1.78 and 1.90, respectively, out of 4), while Trios was the least preferred (mean rank 3.55). Specialists showed significantly stronger relative preference for Smilecloud than laypeople did (rank-biserial r = 0.37, Holm-adjusted p < 0.001), while laypeople were relatively more tolerant of Trios than specialists were (r = -0.44, Holm-adjusted p < 0.001). Within specialists, preference against Trios strengthened significantly with age (Kruskal-Wallis H = 26.2, p < 0.001, ε² = 0.32); no other age or gender association survived correction for multiple comparisons in either group.

Aesthetic preferences for AI-assisted smile designs differed meaningfully between software platforms and, to a lesser extent, between dental specialists and laypeople, with specialists showing a distinctly stronger relative preference for Smilecloud. Age-related differences were most evident among specialists' preferences for Trios. Given the cross-sectional, non-randomized design and the operator-adjusted nature of the outputs evaluated, these findings should be read as associations rather than causal drivers of preference.