AI-designed fixed dental restorations show promise, but expert human oversight remains essential

Limpuangthip N, Hlaing NHMM, Lee SJ, Lee JH. Design quality and time efficiency of AI-based versus human-based design for tooth- and implant-supported fixed dental restorations: A systematic review. Journal of Dentistry, 170, 106644. https://doi.org/10.1016/j.jdent.2026.106644 DESIGN: This systematic review compared fully automated AI-based and human-based digital workflows for the design of fixed dental restorations.

The authors conducted electronic searches in PubMed/MEDLINE, Embase, Scopus, and the Cochrane Library as well as a manual screening of relevant references. The review followed PRISMA guidelines and was based on a registered protocol. To ensure inclusion of the most current evidence, the literature search was performed initially and updated before final data extraction and manuscript preparation. Comparative studies evaluating AI-generated designs for tooth- or implant-supported single or multiunit fixed dental restorations were included when a human-designed restoration served as the comparator. Fifteen studies fulfilled the inclusion criteria. The evaluated restorations included single-unit anterior and posterior crowns, inlays, implant-supported crowns, and three-unit fixed dental prostheses.

Two reviewers independently screened studies, extracted data, and assessed methodological quality using the Methodological Index for Non-Randomised Studies (MINORS). Outcomes included restoration morphology, occlusal relationships, proximal contacts, finish-line detection, marginal adaptation, internal fit, design efficiency, and design success rates. Due to considerable heterogeneity in restoration types, AI software platforms, outcome measures, and assessment methods, a meta-analysis was not feasible, and findings were synthesised narratively.

Across most studies, AI-generated single-unit crowns demonstrated morphology, marginal adaptation, finish-line accuracy, and internal fit that were generally comparable with human-designed restorations. However, AI systems frequently produced less favourable proximal and occlusal contacts and often required refinement by experienced operators to achieve optimal clinical acceptability. AI-designed inlays and three-unit fixed dental prostheses showed greater deviations from human-designed restorations, particularly in occlusal relationships, connector design, and contact formation. Evidence for implant-supported crowns was limited but suggested performance comparable to human designs. AI-based workflows consistently reduced design time compared with conventional human-based workflows and appeared particularly beneficial for less experienced operators.

Current evidence suggests that AI-assisted restoration design can generate clinically acceptable single-unit crown restorations while substantially improving workflow efficiency. Nevertheless, expert human verification and refinement remain necessary, particularly for occlusal function, proximal contacts, aesthetics, and more complex prosthodontic restorations. Current AI models should be regarded as a supportive clinical tool rather than a replacement for professional expertise.