The AI Divide: Uncovering Factors Behind Non-Adoption of AI-powered Health Care in China (Extended Abstract)

The development of artificial intelligence (AI) in healthcare is advancing at a rapid pace. While a growing number of people who embrace AI applications in health-related contexts ranging from consulting diagnostics to seek information for treatment and medication, a significant number of others remain cautious, adopting a more reserved, wait-and-see attitude (Kerstan et al., 2024). What factors that account for such cautious approach to delay the adoption of AI tools? This study investigates how everyday health communication choices reflect broader dynamics of technology adoption and resistance, focusing on individuals who deliberately opt not to use AI tools—such as DeepSeek, Doubao, or ChatGPT—for accessing health information or seeking medical advice.

To contribute to adoption and diffusion research in health contexts (Claudy et al., 2015; Van Oorschot et al., 2018), we examine factors that lead to the delay or resistance in adopting AI as emerging technologies in health communication. Drawing on literature related to the digital divide (Suárez & García-Mariñoso, 2025; Wang et al., 2025), risk perception (Kerstan et al., 2024), and human-AI interaction (Li & Bitterly, 2024), we analyze how key demographics, perceived adoption risks (cognitive, affective, and ethical), social influences, and use of alternative health information channels play a role in shaping the wait-and-see attitudes toward AI. These factors will make a difference in the digital divide in the AI era concerning access, awareness, and literacy.

As part of a large on-going project, we draw insights from four qualitative focus groups conducted in September 2025 (N = 24), which were conducted online, we explored reasons why participants from different cities in China choose not to adopt AI for health-related inquiries. Participants, aged 30 to 59, represented various professions, including manufacturing, public service, and sales. All of them reported limited or no prior experience with AI tools for health queries. Guided by a semi-structured protocol, the discussions focused on health information-seeking behaviors, perceptions of AI from a risk-taking perspective, from which key themes and patterns were identified.

Our findings reveal three primary themes. First, from the perspective of uses of alternative sources, respondents overwhelmingly prefer face-to-face communication with human doctors. One participant remarked, “The doctor can feel my pulse and examine my tongue. AI can’t do any of that, so how can I trust it?” The expertise and authority of human doctors, along with interpersonal trust, significantly influence preferences for in-person consultations and skepticism toward AI.

Second, cognitive, affective, and ethical risks associated with AI tools impact adoption behaviors. Participants expressed concerns about AI misdiagnosing or delaying care, particularly when vague symptom descriptions are involved. There were fears that over-reliance on AI could create a false sense of security, leading to delayed medical attention. Affectively, the absence of empathy emerged as a recurring concern among participants. As one participant noted, “The screen’s answers are cold. AI chatbot makes me more anxious rather than comforted.” Several other participants emphasized the emotional dimension of health concerns, requiring sensitivity and human understanding that AI cannot provide. Additionally, ethical concerns regarding data misuse were prevalent, with one participant stating, “If telecom companies can leak my number, how can I trust them with my medical records?”

Third, the social prevalence of AI affects these seemingly late adopters’ attitudes. As AI becomes more widespread, respondents feel social pressure we’re building for adopting it, especially when in-person consultations in hospitals are time-consuming and costly. In such context, participants generally maintain an optimistic outlook on adopting AI for health-related inquiries in the near future. Additionally, the widespread use of AI among their friends and family, along with their expectations and support, may help respondents overcome both psychological and physical barriers to using AI.

Overall, participants who were non-adopters of AI face disadvantages in accessing AI channels and possess limited literacy in deciding if to use these novel tools. The lack of skills, for example, exacerbates their perception of adoption risks, placing them in a wait-and-see position despite optimistic views about AI’s potential. This dilemma exemplifies the emerging AI-driven digital divide. That is, while AI health technologies are considered as instruments of democratizing accessibility to knowledge, our research indicates that the benefit is unlike to spread equally in society. Late or non-adoption is a manifestation of structural inequality (Celik, 2023; Suárez & García-Mariñoso, 2025; Wang et al., 2025).