This study proposes an LLM-based conversational virtual character intervention paradigm to enhance the conversation experience and emotional regulation of adult ADHD patients. By designing virtual characters and language styles for LLMs, we optimize the emotional resonance and guidance of AI interactions and have developed a software system featuring role-switching, conversation timing, and auto-trigger functionalities. Experimental results demonstrate that this approach improves ADHD patients’ attention focus and emotional regulation, while also increasing user engagement and satisfaction. The study validates the feasibility of conversational virtual characters in ADHD-assisted interventions and provides new directions for future personalized, long-term AI-based interactive interventions.