Artificial intelligence has become a powerful tool for automated dermatological diagnosis, especially for early detection of malignant skin lesions. In this study, a deep learning–based classification framework was developed using dermoscopic images from public datasets. After preprocessing and augmentation, a pretrained CNN model was fine-tuned to distinguish benign and malignant lesions. Performance was evaluated using Accuracy, AUC, Precision–Recall analysis, and confusion matrix. Experimental results show that the proposed model achieves dermatologist-level accuracy, while Grad-CAM visualizations provide explainability by highlighting clinically meaningful lesion areas. The method can be integrated into mobile or tele-dermatology platforms to support faster triage and improve access to skin-cancer screening, particularly in resource-limited regions.