Generative Artificial Intelligence in Dentistry: Clinical Applications, Evidence, Limitations, Ethical Considerations, and a Framework for Responsible Implementation

Generative artificial intelligence (GenAI), including large language models (LLMs) and multimodal foundation models, is increasingly being used in dentistry, dental education, and scientific research. Unlike conventional predictive AI, which is generally developed for defined classification or prediction tasks, GenAI can produce new text, images, summaries, and other outputs from user prompts. In dentistry, potential applications include patient communication, clinical documentation, education, literature synthesis, and selected forms of decision support. These applications are accompanied by important risks, including hallucinated information, bias, privacy breaches, fabricated references, and automation bias. This narrative review summarizes recent evidence on GenAI in dentistry, with emphasis on the distinction between assistive use and autonomous clinical decision-making. Recent systematic, scoping, and narrative reviews describe applications in dental education, oral and maxillofacial surgery, implant dentistry, endodontics, and other areas, but the available evidence remains heterogeneous and prospective clinical validation is limited. Retrieval-augmented generation (RAG) may improve evidence alignment by linking language models to curated external sources, although dental-specific evidence remains early. Based on the reviewed literature, responsible implementation should be task-specific and should include data minimization, verification of outputs, human accountability, documentation, and ongoing monitoring. A seven-stage framework is proposed: define the task, assess risk, protect data, verify output, retain human decision-making, document relevant AI use, and monitor performance. At present, GenAI is best regarded as an assistive technology that can support, but should not replace, professional dental judgment.