The ongoing use of chatbots in healthcare, education, customer service, and mental health has made more apparent the weaknesses of entirely task-focused conversational systems that are non-emotional. Emotion detection has become an essential process of improving human-computer interaction that allows the chatbots to detect the affective states of users and react in a more human-centric and situational behalf. This paper gives a synthesis of the research on emotion-aware chatbot systems and how emotion detection methods, data modalities, and architecture can be used to enhance the quality of interaction. Fifty chosen studies were systematically analyzed to study the trends of publications, prevalent emotion detecting techniques, effectiveness of modality, and system design method. The results show that there is an increasing concentration of quality research in traditional human-computer interaction and artificial intelligence outlets, and there is a growing global concern in the last few years. The use of text-based emotion detection is the most popular in that it is more scalable, whereas the speech, visual, and multimodal detection use more emotion expressiveness and resilience in real life. Multimodal architectures can capture more complex emotional cues better than other electric stimuli, but face difficulties in terms of complexity, privacy and evaluation of the system. The review also shows that most of the current chatbot frameworks are more focused on the technical measures of performance rather than long-term, human-focused evaluation outcomes. In general, the present study provides an insight into the achievements and limitations of the existing research on emotion-sensitive chatbots and emphasizes the necessity to create ethically oriented, culturally sensitive and systematically tested conversational agents in order to promote the development of emotionally intelligent human-computer interaction.