Social media has become an inseparable part of everyday life in modern society to make it easier to interact and communicate with each other. The purpose of this study is to review and compare the deep learning methods implemented in the case of detecting fake news from several previous studies, and to get an overview of the corpus or dataset used by previous studies. This research is also to help researchers identify and map the use of deep learning algorithms in cases of detecting fake news. The research method is conducting a literature survey of 12 literatures obtained from the ScienceDirect and IEEE Xplore websites. The collection of literature that has been surveyed is selected based on the year published in 2021 with the topic of research on detection of fake news using a deep learning approach. The results of this study summarize that the strategy to detect fake news can be done with four approaches, based on the content, based on the writing style, based on the distribution pattern, and based on the credibility of the source. The results of this research also show that the Convolutional Neural Network algorithm is a favorite of researchers by appearing 6 times in the literature collection. The next favorite algorithm is Long Short Term Memory which appears in 5 literatures and Bidirectional LSTM which appears in 4 literatures.