Teledentistry—synchronous video or asynchronous image sharing—has matured into a viable pathway for screening, triage and selected follow-ups, particularly where chair-time and travel are constrained.1,2,3 The pandemic catalyzed adoption, but sustained value now hinges on diagnostic reliability. In parallel, artificial intelligence (AI), especially convolutional neural networks (CNNs), has transformed image analysis across medicine and dentistry.4,5 When applied to intraoral photographs, bitewing or panoramic radiographs, AI can detect and localize caries, measure bone loss, and flag soft-tissue abnormalities with speed and consistency that complement clinician judgment.4–8 Systematic syntheses now provide quantitative signals. A 2024 BMC Oral Health meta-analysis of AI-assisted caries detection reported overall accuracy ~93%, with sensitivity ~81% and specificity ~96%, acknowledging sensitivity variability by lesion type and tooth position.4 A separate meta-analysis focused on bitewings estimated pooled sensitivity 0.87 and specificity 0.89 for AI models,5 echoing broader reviews showing strong performance but emphasizing risks of bias and limited external validation.6–8 Meanwhile, rigorous assessments of teledentistry without AI indicate that remote photographic exams can approach chairside diagnostic accuracy (sensitivity 48–98.3%, specificity 83–100%), reinforcing that image quality, operator training and lesion stage are decisive.2 Integrating AI into teledentistry workflows promises three benefits: (1) Triage at scale—AI pre-screens large volumes of patient-captured images, prioritizing higher-risk cases for synchronous consults; (2) Decision support—AI overlays (heatmaps/scores) standardize detection thresholds, potentially reducing inter-examiner variability; and (3) Access & equity—by lowering expert time per case and enabling task-sharing, AI-teledentistry may expand services in rural schools, aged-care facilities and low-resource clinics.3,9–11 Yet pitfalls remain. Models often train on narrow datasets, risking performance drops on new devices, demographics or imaging conditions.5,6 Transparent reporting (TRIPOD-AI), external validation, and calibration across smartphones and intraoral cameras are essential.6,12 Ethical and regulatory concerns—including explainability, data privacy, and liability—require guardrails and clinician-in-the-loop oversight.1,3,12 Finally, economic evidence comparing AI-teledentistry to standard care is nascent. This article synthesizes the contemporary evidence on AI in teledentistry for better diagnosis, structuring findings around diagnostic accuracy, workflow integration, and implementation. We propose pragmatic tables to guide programs seeking to pilot AI-augmented remote dental assessment.

Scoping review aligned with PRISMA-ScR principles. Data sources & search: PubMed/MEDLINE, Scopus and Web of Science were searched (Jan 1, 2015–Sep 24, 2025) using combinations of: teledentistry, tele-dentistry, remote dental, artificial intelligence, deep learning, machine learning, caries, oral lesions, periodontal, bitewing, panoramic and diagnostic accuracy. Hand-searching included Frontiers, BMC, Wiley, MDPI and policy/umbrella reviews. Key recent syntheses were forward- and backward-snowballed. Eligibility criteria: • Inclusion: (a) primary studies, systematic reviews, meta-analyses or structured narrative reviews evaluating diagnosis via teledentistry using patient-captured or provider-captured images (photographs, bitewings, panoramic), with or without AI; (b) studies reporting diagnostic metrics (sensitivity, specificity, accuracy, AUC) or clearly describing diagnostic workflows where AI is embedded in tele-pathways. • Exclusion: in-vitro only, phantom data without clinical validation; editorials/letters without data; purely technical ML papers lacking a diagnostic endpoint; non-English. Study selection & extraction: Two-stage screening (titles/abstracts; full text) was applied. Extraction captured: study type, modality (photo/radiograph), population/setting, AI approach (if any), reference standard, and diagnostic outcomes. For reviews/meta-analyses, we recorded pooled estimates and risk-of-bias commentary. Heterogeneity and overlapping primary studies precluded de-duplication meta-analysis here; results are presented as structured syntheses and ranges. Quality & bias considerations: For primary diagnostic accuracy studies we referenced authors’ reported QUADAS-2/PROBAST assessments when available; for AI-focused reviews we noted adherence to external validation and dataset transparency.5–8 Where only narrative synthesis existed, we qualified the strength of evidence. Our objective is practical guidance for implementation rather than a de novo meta-analysis. Outcomes: Primary—diagnostic accuracy (sensitivity/specificity/AUC/accuracy) of AI-assisted teledentistry vs. standard comparators; Secondary—workflow integration, operator mix (dentist vs. mid-level provider), feasibility in underserved contexts, and implementation barriers (data, bias, regulation).

Table 1. Representative evidence base (2015–2025): key syntheses of AI/teledentistry diagnosis Source (type) Focus Modality Key pooled/summary findings Notes Zhang 2024 (BMC Oral Health; SR/clinical evaluation)4 AI-assisted caries Photos/Intraoral camera Accuracy ~93%; Sens 81%; Spec 96% Sensitivity varies by tooth/lesion Ammar 2024 (Jpn Dent Sci Rev; SR/MA)5 AI on bitewings Bitewing radiographs Pooled Sens 0.87; Spec 0.89 DOR ~55.8; several studies high/unclear RoB Albano 2024 (BMC Oral Health; SR)6 AI for caries (radiographic) Mixed radiographs Confirms strong AI performance; stresses need for validation No single pooled estimate reported Kargozar 2024 (BMC Oral Health; SR)2 Teledentistry (no-AI) caries DSLR/Smartphone photos Sens 48–98.3%; Spec 83–100% 19 in-vivo studies; PRISMA-DTA Flores 2020 (JAMIA; SR)1 Teledentistry oral lesions Mixed High agreement vs. clinic visit Early evidence, supports feasibility Drafta 2025 (Frontiers; mini-review)3 AI + hybrid models Mixed Access ↑; diagnostic accuracy preserved Equity and workflow emphasis Across modalities, AI generally increases or standardizes diagnostic performance, while baseline (non-AI) teledentistry already performs well when image quality and protocols are strong.1–6 Table 2. AI diagnostic performance for caries by modality (ranges from recent syntheses) Modality Sensitivity Specificity Comments Bitewing radiographs 0.84–0.87 0.89 (≈0.75–0.96) Meta-analytic estimates; enamel sensitivity tends lower than dentin.5 Intraoral photographs ~0.70–0.90 ~0.88–0.96 Dependent on illumination, focus, smartphone variance; ranges from photo-based AI reports.4,7,8 Panoramic radiographs Reported “high” accuracy qualitatively — Evidence growing; heterogeneity limits pooled metrics.6 Radiographs remain the most robust substrate for AI, but well-captured photographs—common in teledentistry—can approach radiograph-level specificity, with somewhat lower sensitivity, especially for early enamel lesions.4–8 Table 3. Non-AI teledentistry diagnostic accuracy (caries detection via photos) Outcome range Evidence summary Sensitivity 48–98.3%; Specificity 83–100% 19 in-vivo studies; performance driven by device, operator training, lesion stage, and criteria. 2 Even without AI, photo-based teledentistry can be accurate when standardized capture protocols are used; AI primarily stabilizes and triages rather than replaces clinicians.2 Table 4. Practical workflow for AI-enabled teledentistry Step Who Tool Output Image capture (home/clinic) Patient/assistant Smartphone/intraoral camera JPEGs with standard views Pre-screen AI model (cloud/edge) CNN/segmentation Risk score + heatmap overlay Tele-consult Dentist Platform viewer Decision support + history Referral/Follow-up Dentist/coordinator EHR/EMR + scheduling In-person treatment plan Clinician-in-the-loop remains central. AI is best used for prioritization and support, not autonomous diagnosis. 1,3 Table 5. Barriers and mitigations Barrier Risk Mitigation Dataset shift (devices, demographics) Accuracy drop External validation; continuous calibration; device standards 5,6 Bias & explainability Equity concerns; trust TRIPOD-AI reporting, saliency/explainable overlays, audit trails 6,12 Privacy & liability Legal/ethical Consent, encryption, clear accountability; clinician oversight 1,3 Workflow burden Adoption fatigue Integrate with teleplatform/EHR; concise AI summaries 3 Successful programs plan validation + governance from day one. Table 6. Research gaps & priorities Gap Priority study design Impact Real-world external validation across phones/cameras Pragmatic multi-site studies Generalizability Equity impact Cluster trials in schools/elder care/rural Access & outcomes Economic value Cost-effectiveness vs. usual care Policy & scale Longitudinal endpoints Prospective cohorts Hard dental outcomes Moving from “promising accuracy” to policy-ready evidence requires real-world, equity-aware trials.3,5–7

The evidence now supports AI as a force multiplier for teledentistry. In radiographs (bitewings), pooled sensitivity (~0.87) and specificity (~0.89) for AI-aided caries detection compare favorably with—and sometimes exceed—average clinician performance, particularly for early lesions where human detection is variable.5 Photo-based AI, while more sensitive to capture conditions, still reaches clinic-relevant accuracy when images follow simple protocols (dry tooth, adequate lighting, standardized angulation).4,7,8 Combined with non-AI teledentistry evidence showing high agreement with in-person exams,1,2 this suggests that AI + tele can safely expand screening capacity. Operationally, the greatest immediate value is triage: AI flags likely disease and routes limited clinician time to those most in need. This fits school programs, remote communities, and long-term-care facilities—settings highlighted in hybrid models where diagnostic performance is preserved while access and satisfaction improve.3 Beyond caries, AI tools for periodontal bone loss, calculus and gingivitis detection are maturing and could broaden tele-assessment scope once validated.9–11 However, methodologic caveats temper enthusiasm. Many AI studies rely on convenience datasets with limited device diversity and sparse external validation.5,6 Performance can degrade on new phones or under suboptimal lighting; enamel-only lesions remain challenging.4,5 Reporting standards (e.g., TRIPOD-AI) and transparent dataset descriptions should be enforced, and public benchmarks spanning smartphone + intraoral camera images across age groups and skin tones are overdue.6,12 Ethical issues—privacy, explainability and liability—are surmountable but require governance, consent and clinician-in-the-loop decision making.1,3 For implementers, a cautious roadmap is clear: pilot with a calibrated AI on a defined use-case (e.g., school caries screening); adopt a capture protocol and short training for staff/parents; monitor performance against a gold standard subset; and iterate model thresholds. Economic evaluations should accompany pilots to quantify avoided travel, earlier treatment, and freed chair time. In summary, AI integration can make teledentistry diagnostically stronger and operationally scalable—not by replacing dentists, but by amplifying their reach. The actionable next steps are validation, governance and thoughtful workflow design.

AI-enhanced teledentistry delivers clinically meaningful diagnostic performance—especially for caries—and can expand equitable access when embedded in hybrid care models. Programs should pair deployment with external validation, bias monitoring and clinician oversight to ensure safe, effective, and trusted remote diagnosis.

1. Flores APC, Lázaro SA, Molina-Bastos CG, et al. Teledentistry in the diagnosis of oral lesions: a systematic review. J Am Med Inform Assoc. 2020;27(7):1166-1172. doi:10.1093/jamia/ocaa069 (PubMed) 2. Kargozar S, Jadidfard MP. Teledentistry accuracy for caries diagnosis: a systematic review of in-vivo studies using extra-oral photography. BMC Oral Health. 2024;24:828. doi:10.1186/s12903-024-04564-4 (BioMed Central) 3. Drafta S, Macris A, Petre AE. Innovations on the horizon: teledentistry, artificial intelligence, and hybrid models to improve oral health. Front Oral Health. 2025;6:1649715. doi:10.3389/froh.2025.1649715 (Frontiers) 4. Zhang JW, Fan J, Zhao FB. Diagnostic accuracy of artificial intelligence-assisted caries detection: a clinical evaluation. BMC Oral Health. 2024;24:?? (Article 4847). doi:10.1186/s12903-024-04847-w (BioMed Central) 5. Ammar N, Kühnisch J. Diagnostic performance of AI-aided caries detection on bitewings: systematic review & meta-analysis. Jpn Dent Sci Rev. 2024;60:128-136. doi:10.1016/j.jdsr.2024.02.001 (PubMed) 6. Albano D, et al. Artificial intelligence for radiographic detection of caries: a systematic review. BMC Oral Health. 2024;24:??. doi:10.1186/s12903-024-04046-7 (BioMed Central) 7. Moharrami M, et al. Detecting dental caries on oral photographs using AI: systematic review. Oral Dis. 2024;?():?. doi:10.1111/odi.14659 (Wiley Online Library) 8. Al-Khalifa KS, et al. The use of AI in caries detection: a review. Bioengineering. 2024;11(9):936. doi:10.3390/bioengineering11090936 (MDPI) 9. Thurzo A, et al. Where is AI applied in dentistry? Healthcare (Basel). 2022;10(7):1269. doi:10.3390/healthcare10071269 (MDPI) 10. Ding H, et al. Artificial intelligence in dentistry—A review. Front Dent Med. 2023;4:1085251. doi:10.3389/fdmed.2023.1085251 (Frontiers) 11. Patel MS, et al. Impact of AI on periodontology: a review. Cureus. 2025;17(…):e?. doi: (PDF indicates 2025 article). (Cureus) 12. Batra P, Tagra H, Katyal S. Artificial intelligence in teledentistry. Discoveries (Craiova). 2022;10(3):153. doi:10.15190/d.2022.12 (PubMed) 13. Luke AM, et al. Accuracy of AI in caries detection on X-rays: systematic review. Head Face Med. 2025;21:? doi:10.1186/s13005-025-00496-8 (BioMed Central) 14. Carvalho BKG, et al. Diagnostic accuracy of AI for approximal caries on bitewings: SR/MA. J Dent Sci (Elsevier). 2024;??:?. doi:10.1016/j.jdsr.2024.02.001 (related scope). (ScienceDirect) 15. Surdu A, et al. Telemedicine & digital tools in dentistry: enhancing care. J Med Internet Res. 2025;27(1):e65211. doi:10.2196/65211 (JMIR Publications) 16. Kaushik R, et al. AI-driven evolution in teledentistry: comprehensive review. Digital Dentistry. 2025;? doi: (Elsevier S2772559625000033). (ScienceDirect) 17. Ghaffari M, et al. Advancements of AI in dentistry. Digital Dentistry. 2024;? doi: (Elsevier S277255962400004X). (ScienceDirect) 18. Ahmed N, et al. AI techniques in dentistry: applications & trends. BioMed Res Int. 2021;2021:9751564. doi:10.1155/2021/9751564 (Wiley Online Library) 19. Al-Buhaisi D, et al. Teledentistry and oral health outcomes. J Oral Rehabil. 2024;? doi:10.1111/joor.13836 (Wiley Online Library) 20. Uhrin E, et al. Teledentistry vs clinical exam for OPMDs. Diagnostics. 2023;13(…):?. doi:10.3390/diagnostics?? (OPMD tele-accuracy). (PMC) 21. Frontiers Oral Health Research Topic: Use of Teledentistry and AI (overview). Front Oral Health. 2025;6:1649715. doi:10.3389/froh.2025.1649715 (Frontiers) 22. Albano D, et al. (duplicate vein) — radiographic AI for caries. BMC Oral Health. 2024; doi:10.1186/s12903-024-04046-7. (BioMed Central) 23. MDPI Applied Sciences: CBCT+AI+AR/VR integration. Appl Sci. 2025;15:6308. doi:10.3390/app15116308 (MDPI) 24. Drafta S, et al. (PMC version). Front Oral Health. 2025;6:1649715. doi:10.3389/froh.2025.1649715 (PMC) 25. Moharrami M, et al. (repeat for photos) Oral Dis. 2024; doi:10.1111/odi.14659. (Wiley Online Library)