Appraising Artificial Intelligence Studies in Dental and Maxillofacial Imaging: A Task-Routed Framework for Reporting Completeness and Risk of Bias

Artificial intelligence studies in dental and maxillofacial imaging use different designs and endpoints, but reviews often apply a single adapted appraisal instrument across them.Recent updates to reporting guidelines and risk-of-bias tools have clarified their respective scopes.This narrative methodological review proposes an appraisal framework in which the review question, study design, and each eligible evaluation determine the instrument used.Diagnostic accuracy estimates are assessed using appropriate criteria, with comparative accuracy evaluations assessed separately according to relevant methodological guidance.Diagnostic or prognostic multivariable models are evaluated for risk of bias and methodological quality, with model development quality and evaluation risk of bias reported separately.Structural segmentation, localization, registration, and image-enhancement evaluations receive a declared descriptive appraisal where the framework provides no established task-specific risk-of-bias route.Reporting completeness is assessed separately using the relevant reporting guidelines.Hybrid reports may contribute to more than one route, with evaluations and estimates linked to the underlying study and dataset.Results are presented by route, with explicit denominators, rather than combined into a single quality score.The proposed extraction items address annotation provenance, patient-level separation between data partitions, within-patient clustering, dataset reuse, clinical exclusions, and modality-specific acquisition characteristics.The review also identifies limitations concerning general-purpose multimodal models and certainty-of-evidence assessment.The framework offers practical guidance and a proposed minimum reporting set for authors, reviewers, and editors.Its routing reliability has not been evaluated; therefore, the framework is intended to improve transparency and comparability and requires empirical testing.