Background: Artificial intelligence (AI), particularly convolutional neural networks (CNNs) and vision transformers (ViTs), is transforming dental radiographic diagnosis across multiple clinical domains. Methods: A narrative review searched PubMed, PubMed Central, BMC Oral Health, and Google Scholar for systematic reviews, meta-analyses, and original studies published between 2019 and 2025, covering ten domains: caries detection, periapical lesion identification, alveolar bone loss, peri-implant assessment, tooth detection and segmentation, cone-beam computed tomography (CBCT) analysis, vertical root fracture (VRF) detection, cephalometric landmark analysis, dental anomaly detection, and AI-versus-clinician comparison. Results: Pooled sensitivity was 0.94 and specificity 0.91 for caries detection on bitewing radiographs. Periapical AI sensitivity was 92.9% with specificity of only 58.6%, yielding a positive predictive value of 15.3%. Peri-implant bone loss assessment achieved AUC of 0.94. Vision transformers outperformed CNNs in 58% of dental imaging studies. VRF detection accuracy ranged from 75% to 97.8%. Cephalometric landmark detection rate reached 98.29%. AI diagnosed radiographs in 1.5 seconds versus 53.8 seconds for clinicians. ANSI/ADA Standard 1110-1:2025 provides the first formal dental AI validation framework. Conclusions: AI achieves substantial and reproducible diagnostic accuracy across dental radiographic tasks. High false-positive rates, demographic dataset bias, limited multi-centre external validation, and regulatory gaps must be resolved before broad clinical implementation.