This work presents a preliminary exploration into the development and testing of a bio-adaptive Virtual Human (VH) aimed at reducing users’ arousal through personalized relaxation protocols. The VH integrates (i) large language model–based personalization, and (ii) real-time physiological feedback using electrodermal activity (EDA). Three experimental conditions were tested: (1) a standard protocol with predefined, non-personalized content; (2) a personalized protocol generated through a preliminary user survey; and (3) a personalized protocol further adapted in real time using EDA-driven neurofeedback. Participants’ arousal was assessed using both subjective reports and physiological indicators. Although overall results suggest a general tendency toward reduced arousal, statistical significance was not achieved, likely due to the study’s exploratory nature and limited sample size. Notably, some participants reported discomfort when encountering personal details in the relaxation protocol, indicating that how individualized content is introduced must be carefully managed. This preliminary work highlights the potential of combining conversational AI, neurofeedback, and psychophysiological monitoring but also underscores the importance of user-centered design and careful personalization strategies for future implementations.