Artificial intelligence in clinical dentistry

Abstract

Background

Artificial intelligence (AI) is increasingly integrated into clinical dentistry, yet ethical, legal, and implementation uncertainties continue to limit its widespread adoption. This study aimed to generate a prioritized, hypothesis-generating item bank on ethical principles, implementation barriers, and governance frameworks for AI in dentistry using a modified Delphi approach.

Methods

A 2-round exploratory modified e-Delphi study was conducted from January through April 2026. Round 1 (n = 12) used 6 open-ended questions to generate items through thematic analysis. Round 2 (n = 9) required participants to rate 109 items across 12 domains using a 10-point Likert scale. Prespecified criteria for item agreement were defined a priori as a median score 7 or greater and an interquartile range 2 or less.

Results

Of the 109 items, 84 (77.1%) met the agreement criteria. Complete agreement (100%) was observed for both new ethical principles (commercial bias, epistemic integrity) and all items related to structured ethical decision making in AI-clinician tensions. High levels of agreement were observed for shared responsibility model, mandatory AI training, explainable AI, privacy-by-design systems, and automated audit trails. Lower agreement was observed for informed consent delivery formats as well as several disclosure and documentation-related items. Overall, 74% of items were rated as top priority by this panel, whereas 25% were considered important but controversial.

Conclusions

The findings identify priority areas for future research, while also highlighting areas of persistent controversy that require further empirical validation. These results should be considered provisional and require confirmation in larger, more representative Delphi studies before informing clinical guidance or policy.