Artificial Intelligence–Driven Smile Design in Prosthodontics: A Systematic Review and Meta-Analysis of Data Modalities, Validation and Clinical Utility

Purpose Artificial intelligence (AI) has been suggested as an additional tool for diagnosing, treating and communicating with patients in prosthetics and other areas of medicine. This review aims at reviewing the current literature on the types of data used by such systems, validation techniques and clinical utility. Study selection Electronic searches were conducted on the PubMed/MEDLINE, Cochrane Library, Dimensions, and Google Scholar databases. Included studies evaluated AI-assisted tools for esthetic smile analysis, smile simulation, shade recommendation, or digital comprehensive planning in adults. Data regarding study design, data modalities, validation comparators, esthetic and functional outcomes, patient and professional-rated outcomes, and workflow metrics were abstracted. Results The evidence base comprised nine studies: technical validation studies, cross-sectional preference studies, and non-randomized clinical or cohort studies. No randomized controlled trial was identified. The meta-analysis indicated a high overall rate of positive responses or clinical success of 0.84 (95% CI 0.72–0.91). For continuous outcomes, results favored AI-assisted workflows, with a pooled standardized mean difference (SMD) of 1.29 (95% CI 0.02–2.56). Validation reliability was high, with a pooled correlation of 0.92 (95% CI 0.74–0.98). Individual studies also showed high repeatability for 3D AI tasks (ICC 0.973–1.000) and improved shade recommendations (ΔE00 ≤ 1.8: 85% vs 70%; mean ΔE00 1.42 ± 0.56 vs 1.88 ± 0.72; time 2.9 ± 0.8 vs 4.6 ± 1.2 min). Pooled estimates came from only three or four studies each, and there was notable between-study heterogeneity (τ² 0.261–1.166) and wide confidence intervals, with the pooled SMD not being clearly differentiated from the absence of effect. Conclusion With regard to a small set of evidence that is not randomized or long-term, the use of digitally assisted smile design systems was shown to be positively effective on the measurement accuracy, shade matching and workflow efficiency (moderate certainty), while its effects on esthetic preference and patient satisfaction were contradictory (low certainty). These results are encouraging yet preliminary, and they support the use of digital smile design systems as an add-on in prosthodontic work-flows.