The rapid proliferation of smart fitness wearables has revolutionized personal health monitoring, yet understanding the determinants of post-adoption behavioural intentions remains underexplored. This study employs the Expectation-Confirmation Model (ECM) as a theoretical lens to investigate key post-adoption factors influencing continued usage of wrist-worn fitness devices. Using a qualitative research approach, Focus Group Discussions (FGDs) were conducted with 34 active users categorized based on device usage duration. The study leverages GPT-4, a Large Language Model (LLM), for automated thematic extraction and behavioural analysis, to enhance classification accuracy and scalability. Findings reveal that confirmation of expectations, perceived usefulness, and user satisfaction significantly drive long-term engagement, while hedonic motivation and user interface emerge as critical extensions to the ECM framework in this domain. By integrating AI-driven qualitative analysis with theoretical models, this research offers novel insights for wearable technology developers, emphasizing the need for usability enhancements, gamification features, and personalized health tracking to improve user retention. The study also underscores the advantages of LLM-assisted qualitative research in improving classification efficiency and scalability over traditional manual methods. Future research should further refine AI-driven analytical models and explore additional variables such as privacy concerns and personalization to deepen the understanding of wearable technology adoption.