Chatbots are dialogue systems driven by Natural Language Processing (NLP) and Artificial Intelligence (AI), extensively utilized in areas like customer support, education, and healthcare. Nonetheless, the variety in approaches to chatbot development, ranging from rule-based systems to generative AI, creates difficulties in harmonizing design decisions with user requirements and technical limitations. This research seeks to examine and contrast the primary approaches employed in chatbot creation: rule-based, retrieval-based, and generative-based systems. Employing a descriptive-qualitative methodology, the study is carried out in the first quarter of 2025 and utilizes scholarly literature, technical documents, and case studies of Mitsuku, Google Assistant, and ChatGPT, concentrating on applications in Indonesia and Malaysia. A comparative analysis assesses every method based on development complexity, accuracy, flexibility, user experience, interpretability, cost, and ethical risks. The results indicate that rule-based systems provide low expenses and significant transparency, but they fall short in scalability and flexibility. Retrieval-based systems excel in accuracy for domain-specific queries but struggle with new interactions. Chatbots based on generative models provide the most natural and contextually aware interactions, but they require significant resources and present issues related to interpretability and ethics. The research suggests that hybrid models integrating control and adaptability could be the most efficient approach. Further studies are required to improve transparency in generative systems, reduce bias, and create adaptive hybrid architectures appropriate for Southeast Asian use.
Yani et al. (Wed,) studied this question.