Online social platforms provide people with a new way of communication and exchange, and also make the rapid spread of fake news possible. Therefore, automatic detection of fake news has become crucial. Most existing fake news detection methods use deep learning models based on news content and social context. However, fake news often has strong misleading characteristics and can trigger users’ high emotions and questioning stances, which can provide useful information for fake news detection. In this paper, we propose a novel fake news detection model that integrates Comment Emotion and User Stance, namely CEUS. Specifically, We first train the emotion recognition task and fake news detection task jointly by multi-task learning, which extracts high-dimensional representations of emotional features from news comments. Then, based on the semantic representations of news content and comments, we calculate their pairwise similarity and take into account the temporal nature of user comments to obtain the forward and backward user stance. Finally, we concatenate emotion features and stance features to detect fake news together. Extensive experiments on real-world fake news datasets Weibo-16 and Weibo-20 show that the CEUS outperforms the state-of-the-art methods, and learns useful emotion features and stance features to efficiently improve the performance of fake news detection.