Generative Adversarial Networks (GANs) are used in several applications of underrepresented domains. The GAN’s ability to generate data without explicitly modeling the probability distribution enables the synthesis of unlabeled samples and effectively imposes a higher-order consistency. Medical data is costly to aggregate and, in many cases, can risk the patient’s privacy. The high fidelity of generated data by GANs has led many researchers in healthcare to use GANs to accommodate the scarce datasets in the domain. Coronavirus disease 2019 (Covid-19) constitutes one of those cases. As a virus that appeared two years ago, limited information and data exist compared to other phenomena present for decades. Thus, Covid-19 datasets are imbalanced and tend to be biased towards negative Covid-19 cases, as usually, only less than 10% of the data belong to a Covid-19 positive class. This study conducts a detailed review of all eleven existing GAN applications that aid Covid-19 detection. The existing work focuses on the synthesis, segmentation, classification, and saliency map generation of radiological data namely Chest X-rays (CXR) and Computed Tomography (CT). Both imaging data have been reliable information sources for detecting Covid-19. Furthermore, the application of GAN in Covid-19 spread modeling with data assimilation and uncertainty quantification is also explored. We believe that the illustrated findings will provide a clear overview and inspire future innovations in the right direction.