Prehypertension, an early stage in the development of hypertension, impacts a substantial segment of the adult population worldwide. Addressing this issue, our study introduces HabitBot, an AI-driven chatbot tailored to encourage physical activity (PA) habits among individuals with prehypertension. HabitBot combines natural language processing with multidisciplinary approaches, drawing from both theoretical frameworks and empirical studies. The chatbot development followed a systematic, five-phase process: comprehensive needs assessment, literature review on behavior change theories, analysis for selecting effective behavior change techniques (BCTs), prototype design through intervention mapping, and refining the intervention based on user feedback. The outcome includes a prototype that integrates the Health Action Process Approach and Habit Formation Theory, utilizing twelve identified BCTs effective in fostering PA habits. User feedback further refined the chatbot across multiple dimensions such as user interface, content accessibility, and privacy. HabitBot exemplifies an innovative integration of behavior change strategies with advanced language model technology, paving the way for digital health interventions in chronic disease prevention. Future studies should assess its long-term efficacy in habit formation and explore its applicability to various demographic groups.