With the rapid advancement of artificial intelligence, deep learning-based multimodal data fusion technology has found increasing applications in the field of dental and stomatological diseases, and it has demonstrated great potential. This technology constructs deep neural networks to automatically extract and fuse features from multimodal data, such as clinical text information, medical imaging, intraoral and extraoral digital scanning records, and histopathological data. By fully leveraging the complementary information among these diverse data sources, it substantially enhances the accuracy of oral disease diagnosis, the rationality of treatment planning, and the reliability of prognosis prediction. This survey systematically explores the core advantages of deep learning in multimodal data fusion; outlines key main strategies, such as early fusion, intermediate fusion, and late fusion along with their applicable scenarios; and summarizes the specific applications and latest research progress of this technology in oral diseases, including oral cancer, periodontitis, dental caries, and temporomandibular disorders. We also highlight critical challenges in its clinical translation, such as complexity in multimodal data alignment, scarcity of high-quality annotated data, insufficient model interpretability, and privacy and security issues. In the future, this technology is expected to develop toward new learning paradigms, intelligentization, and adaptive fusion while integrating explainable artificial intelligence to enhance clinical credibility. Ultimately, large-scale, multicenter clinical validation systems will be required to ensure its safe and effective deployment in clinical practice.