Answers are grounded in this paper and cite their sources.
Artificial intelligence-based systems (AIS) are becoming part of dentistry. Their integration raises unresolved tensions for users such as trust, explainability, responsibility, professional autonomy and real-world usefulness. This concise review examines these tensions and discusses how they may shape the adoption of AIS in dentistry. This concise review was designed as a narrative review. An initial screening suggested that challenges associated with the implementation of AIS in dentistry appear to remain insufficiently explored in the dental literature. An author with expertise in marketing then formulated a list of AIS implementation issues considered well established in marketing research, drawing particularly on reference publications from leading journals. Based on these issues, the authors identified a set of tensions, and the dental authors translated them into concrete examples and implications for dental implementation. In dentistry, these tensions arise at three interconnected levels. At the human level, AIS may support diagnosis, risk communication and shared decision-making, but may also foster over-reliance or unjustified rejection. At the human–AIS interface, tools can flag lesions, prioritise cases or automate workflows that redistribute agency and require clear boundaries between assistance, delegation and responsibility. At the AIS level, dental applications face persistent issues of generalisability, privacy, adaptability and validation across populations, devices and care settings. The value of AIS in dentistry will depend less on technical performance alone than on the capacity to design, evaluate and implement user-centered tools with a co-designed approach involving all stakeholders, from patients and clinicians to institutions and decision-makers. Anticipating these tensions may help support trustworthy, context-sensitive and sustainable AIS adoption in oral health care.
No takes yet. Share an insight, caveat, or question.
Ducret et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: