This study analyzed to what degree artificial intelligence conversation systems can recognize and respond to human’s emotion language, and it aimed to suggest areas that need to be supplemented for emotional exchange conversations. For this purpose, the study conducted experiments on three kinds of speakers with artificial intelligence conversation system. As a result, the following points have to be solved for the emotional exchange in the artificial intelligence conversation system. First, from the viewpoint of recognition, it is necessary to strengthen the learning of various phrases and expressions related to the emotion, since only the emotion language composed of one word can be grasped at this time. Second, from the viewpoint of the reaction, it is necessary to construct effective dialogue and content by learning the proper dialogue principles as a communication partner for humans that can empathize and communicate, rather than as a machine that evokes instrumental purpose. The significance of the study is what an artificial intelligence conversation system needs for emotional exchange with humans to establish a foundation for a more accurate and efficient development of technology.