The zoonotic viral disease monkeypox has, of late, taken its stage as a global health issue, associated with outbreaks reported outside its endemic regions. Early disease detection is the most critical step in controlling this disease’s spread, but traditional approaches, including PCR, are too slow and cost prohibitive besides lacking in most settings. The deep learning framework allows for a scalable approach: fast, image-based diagnostics of diseases such as monkeypox from images of skin lesions. There is a huge research gap because of limited access to diversified extensive datasets, where failure occurs when current models fail to generalize to another population. Besides, deep learning models lack interpretability, which is a significant challenge for clinical adoption. This study reviewed the application of deep learning techniques, including CNNs, DenseNet, transformer-based architectures, attention mechanisms, and hybrid models, for the detection of monkeypox. The study provides an overview of 18 papers where those approaches are discussed. Found results include the classification of monkeypox lesions by CNNs and DenseNet models with an accuracy level of up to 95.18%. An attention mechanism and adversarial training improve the performance, but issues with the diversity of the dataset and computational cost persist. The impact of these changes is considerable in public health since artificial intelligence-based tools may help to better and faster detect monkeypox, especially in resource-poor settings.