Early detection of skin diseases, which helps guide treatment decisions and ultimately improve patient outcomes, is critical. With rapid developments in artificial intelligence (AI) and deep learning, automated diagnostic systems are increasingly able to match, if not exceed, expert dermatologists in diagnostic accuracy. This review provides an extensive overview of the major algorithms being used for early detection of skin disease, from classical proven machine learning approaches to advanced Convolutional Neural Networks (CNNs). This includes increasing usage of hybrid and ensemble models, along with newer methods such as attention mechanisms and explainable AI. We detail the key benchmark datasets, evaluation methodologies, and comparative performance. We also discuss critical emerging challenges, including poor data diversity, class imbalance, and lack of consistent clinical performance. Finally, we look at new trends and forward-looking directions aimed at developing more robust, reliable, and clinically useful diagnostic systems