This paper explores the incremental development of PLEA, a humanoid intelligent virtual agent with emotional capabilities, that resides in virtual and physical environments. PLEA's evolution encompasses advanced emotion recognition and response generation. The study presents a survey that evaluates PLEA's ability to foster intuitive and engaging human-robot interactions. Findings highlight mixed user perceptions, with participants acknowledging PLEA's moderate ability to mimic human emotions but describing interactions as somewhat artificial. Significant differences in perceptions between students and professionals were noted, with the latter group feeling more relaxed during interactions. The paper discusses key challenges, such as achieving seamless anthropomorphism while avoiding the “uncanny valley” phenomenon and emphasises the need for iterative design improvements based on user feedback. By advancing PLEA's emotional intelligence, this research aims to enhance human-robot interaction, paving the way for applications in education, therapy, and customer service while addressing critical ethical considerations.