Researchers from the University of Surrey have issued a critical assessment of artificial intelligence chatbots in national health services, warning that aggressive messaging and excessive anthropomorphism may deter patients from booking essential medical appointments. The study, published in the journal Lingua, examines patient interactions with Asa, a generative AI receptionist deployed by a general practice in Islington, North London, to facilitate cervical screening bookings via WhatsApp. The investigation combined qualitative interviews with clinic patients and a quantitative survey of three hundred individuals eligible for NHS cervical screening. Findings indicate that conversational design carries as much weight as technical functionality. Participants responded favorably to a friendly, choice-oriented tone that integrated seamlessly into daily routines. Several patients noted that engaging with a named, female-presenting AI lowered barriers to discussing sensitive health matters, such as menstrual cycle conflicts, compared to traditional telephone receptionists. Conversely, the research identified significant friction points that risk patient disengagement. Follow-up messages dispatched within twenty-four hours were widely characterized as intrusive, while imperative phrasing was perceived as aggressive rather than supportive. Patients managing mental health conditions, neurodivergence, or substantial caregiving duties reported that the implicit pressure to respond rapidly felt unfair and counterproductive. Ethical considerations emerged as the most consistent area of concern. Participants expressed apprehension regarding data privacy, potential impersonation, and the blurring of boundaries between automated systems and human clinicians. The chatbot prompt inviting users to interact as if speaking with a real person was frequently cited as a negative factor, interpreted as deceptive rather than reassuring. Dr. Doris Dippold, lead author and Associate Professor in Intercultural Communication at the University of Surrey, emphasized that human-like design traits are not universally beneficial in clinical contexts. She noted that while anthropomorphism can establish initial rapport, it undermines institutional trust when it conflicts with patient expectations for transparency and accountability. The study addresses a critical public health challenge, as cervical screening participation across the United Kingdom declined by 5.3 percent during the 2023 to 2024 period. Demographic disparities remain pronounced, with ethnic minority and socioeconomically disadvantaged populations consistently underrepresented in uptake statistics. The Islington clinic, situated in a highly diverse and deprived community, highlights the necessity for accessible and culturally competent digital health tools. Based on the findings, the research team outlines fundamental design principles for medical AI communication. Health service chatbots must prioritize respectful, transparent, and emotionally supportive interactions. Dr. Dippold stressed that emotional validation is a clinical necessity rather than an optional enhancement. When algorithmic interfaces fail to establish credibility or adopt a coercive tone, patients are likely to disengage entirely, exacerbating existing gaps in preventive healthcare access. The findings signal a shift in digital health strategy, moving away from purely transactional automation toward conversational models that explicitly acknowledge patient autonomy, data boundaries, and psychological safety. As national health systems increasingly integrate generative AI into administrative workflows, aligning communication design with patient-centered ethics will be essential to maintaining clinical engagement and improving preventive care outcomes.