This paper presents potential of gastritis images generated by generative adversarial networks (GANs) for gastritis classification. GANs are popular methods that produce novel samples from high-dimensional data distribution, such as images or sounds. It has been reported that GANs-generated images are useful for improving classification tasks. Since collecting medical image data is difficult compared to natural images, it would be helpful if we can use GANs-generated medical images. Hence, we generate gastritis and non-gastritis images via adversarial learning and use generated images for supervised recognition tasks for evaluating their effectiveness of the gastritis classification.