Who Trusts AI for Health Information? A Cross-National Survey on Trust Determinants in Four European Countries.

Abstract

As generative artificial intelligence (AI) systems such as ChatGPT become increasingly popular sources of health information, understanding what shapes users' trust in AI-generated health information (AI-HI) is essential. Despite growing use, little is known about how human- and source-related factors jointly influence trust across national contexts. Drawing on frameworks from AI trust and online health information research, this study used multi-group structural equation modeling with representative samples from Austria (N = 502), Denmark (N = 507), France (N = 498), and Serbia (N = 483) to predict trust in AI-HI and its effect on intention to use AI-HI. AI literacy and performance expectancy consistently increased trust across countries, while social norms and prior AI-HI experience showed smaller, context-dependent effects. Health literacy, personal innovativeness, effort expectancy, and surveillance risk perceptions were not significant. Informational risk perceptions had only a weak negative effect on trust, indicating that while concerns about inaccuracy can reduce confidence, they play a relatively minor role in shaping it. Trust strongly predicted intention to use AI-HI in all countries, with path-level effects largely stable across contexts. These findings suggest that trust in AI-HI is shaped more by digital capabilities, perceived utility, and social endorsement than by privacy concerns or health literacy. Future research should examine how digital literacy interventions and transparency standards can foster informed trust in these systems.