This study proposes a generative AI-supported workflow for early-stage chatbot UI design by integrating ChatGPT and MidJourney into the ideation process. While existing approaches often emphasize iterative testing and participatory design, there remains a lack of structured tools that support multimodal creativity in the early phases of chatbot development. To address this, we analyzed 80 real-world chatbot interfaces across various industries using a dual-layered taxonomy: visual form (abstract, cartoon, human-like) and sector classification based on ISIC standards. Findings revealed consistent correlations between chatbot form, emotional tone, and interaction strategies across domains.