This study systematically reviews 55 landscape studies that use computer vision methods to interpret social media images and summarizes their spatiotemporal distribution, research themes, method trends, platform and data selection, and limitations. The results reveal that in the past six years, social media–based landscape studies, which were in an exploratory period, entered a refined and diversified phase of automatic visual analysis of images due to the rapid development of machine learning. The efficient processing of large samples of crowdsourced images while accurately interpreting image content with the help of text content and metadata will be the main topic in the next stage of research. Finally, this study proposes a development framework based on existing gaps in four aspects, namely image data, social media platforms, computer vision methods, and ethics, to provide a reference for future research.