Dentists Warmed to AI for Oral Lesion Detection, but Attitude Carried the Most Weight
Artificial intelligence is moving closer to the dental chair, and this survey suggests dentists are already leaning toward it when the task is early recognition of oral potentially malignant disorders (OPMDs). In a 500-dentist cross-sectional study from Egypt, behavioral intention to use AI-assisted clinical decision support was shaped most strongly by attitude toward AI, with trust, perceived usefulness, perceived ease of use, and AI-assisted clinical decision reliance all contributing to the model.
The hierarchy matters. When clinicians are asked whether they would actually use a new decision-support tool, favorable feeling toward the technology often becomes the bridge between technical performance and routine adoption. In this dataset, attitude toward AI showed the strongest association with behavioral intention, with a standardized coefficient of 0.550. AI-assisted clinical decision reliance followed at 0.220, then perceived usefulness at 0.180, trust at 0.140, and perceived ease of use at 0.120.
Why OPMD Detection Is a Natural Test Case for AI Adoption
OPMDs occupy a difficult clinical space. Lesions such as leukoplakia, erythroplakia, oral lichen planus, and oral submucous fibrosis can look subtle at first glance, yet they may precede oral squamous cell carcinoma. The burden of oral cancer remains substantial, and delayed recognition still pushes many patients into later-stage disease, where treatment becomes more complex and outcomes worsen.
That makes the dental setting an unusually important proving ground for AI. Dentists are often the first clinicians to inspect oral mucosa during routine care, so a decision-support system that can flag suspicious lesions early could shorten the path from exam chair to biopsy or specialist referral. But the path from algorithm to practice is not governed by performance metrics alone. It depends on whether dentists trust the system, find it easy to use, and believe it improves their clinical judgment rather than competing with it.
A single high-performing model can still sit idle if clinicians see it as opaque or disruptive.
What the Extended Technology Acceptance Model Captured
Investigators used an extended Technology Acceptance Model to study how dentists think about AI-assisted decision support for OPMD detection. The framework included perceived ease of use, perceived usefulness, trust in AI, attitude toward AI, behavioral intention to use AI, and a newly developed construct labeled AI-assisted clinical decision reliance. That last element is especially relevant in oral medicine, where a suspicious lesion may prompt a dentist to pause, consult an imaging aid, and decide whether surveillance or biopsy is warranted.
The survey was administered cross-sectionally to licensed dentists practicing in Egypt across academic institutions, government facilities, private practices, and professional networks. The final analytic sample reached 500 dentists, far above the minimum regression-based sample size estimate and large enough to support confirmatory factor analysis, structural equation modeling, and bootstrap mediation testing. The questionnaire itself contained 35 items spanning demographic data, AI awareness and exposure, the core TAM domains, and the reliance construct.
The measurement properties were described as generally satisfactory, and the overall model fit was good. In practical terms, that suggests the constructs held together well enough to support the structural analysis rather than dissolving into noisy, overlapping survey responses. The model explained 67.5% of the variance in behavioral intention, a substantial proportion for a behavioral adoption framework.
Trust, Usefulness, and the Bedside Logic of Adoption
Trust in AI often becomes the quiet determinant in clinical adoption studies. Dentists do not hand over responsibility to software, but they do need confidence that an algorithm’s recommendation is grounded in evidence, reproducible, and understandable in the context of a real patient’s lesion. Explainability becomes more than a technical feature; it is the mechanism that lets clinicians decide when to follow a recommendation, when to challenge it, and when to seek tissue diagnosis.
Perceived usefulness also carries obvious clinical weight. If AI can help dentists recognize lesions that are easy to miss during a busy examination, or if it can strengthen triage decisions for referral and surveillance, its perceived value rises quickly. Ease of use matters as well, but it tends to function as an entry point rather than the final reason for adoption. A tool that is simple but clinically uninspiring will not survive long in a workflow already burdened by documentation, imaging, and patient counseling.
The findings also suggest an indirect route through attitude. Significant indirect effects operated through attitude toward AI, meaning perceptions of usefulness, trust, and usability may shape intention partly by changing how dentists feel about the technology overall. In implementation terms, education and interface design may not be enough on their own. They have to make the clinician comfortable enough with the technology to turn curiosity into routine use.
Why the Study Still Leaves the Real-World Question Open
This was a cross-sectional survey, so it captures readiness, not behavior. Behavioral intention is an important step, but it is not the same as actual use in daily practice, where time pressure, case complexity, medicolegal caution, and institutional support can change decisions quickly. The study also measured perceptions in a single country and did not test whether enthusiasm differs across practice settings, age groups, specialties, or exposure to prior AI tools.
Even so, the message is concrete. Dentists surveyed here were broadly favorable toward AI-assisted decision support for OPMD detection, and the strongest lever was attitude. Future deployment will depend on systems that are explainable, easy to integrate into a real appointment flow, and credible enough to earn trust during high-stakes lesion assessment.
Longitudinal work will have to answer a tougher question: whether positive intention in a survey room becomes sustained use when a suspicious oral patch appears in front of the dentist chair.