Accurate early screening of oral cancer together with its precursor lesions remains essential to obtain better treatment results. Current histopathological procedures require vast amounts of time for analysis while remaining subject to different interpretation results from different specialists. This paper provides an extensive examination of artificial intelligence methods for automatic diagnosis of oral cancer and precancerous lesions. The paper explores machine learning (ML) and deep learning (DL) models together with their training protocols for recognizing malignant and premalignant cellular components. The results from experimenting show that AI-based systems provide outstanding diagnostic results which suggest their value for future clinical deployment. The research presents an overview of future prospects and difficulties that involve data quality enhancement together with model interpretation methods and clinical implementation strategies. © 2025 IEEE.