Understanding healthcare professionals’ ethical perspectives on AI-simulated compassion in healthcare: a qualitative study

As artificial intelligence (AI) continues to advance in emotional support and communication in healthcare, AI-simulated compassion, defined as an AI-enabled approach that uses algorithms to recognize and respond to patients’ emotions or suffering and generate supportive communication, has attracted increasing attention. However, how healthcare professionals understand such responses and their potential implications for care relationships, professional roles, and ethical responsibilities remain insufficiently explored. This study aimed to explore healthcare professionals’ understanding and anticipatory ethical judgments regarding the potential clinical application of AI-simulated compassion, with particular attention to the contextual appropriateness of AI-generated supportive responses, emotional support tasks and professional care relationships, and potential risks and responsibilities. A qualitative descriptive study was conducted using purposive and maximum variation sampling. Twenty healthcare professionals were recruited from public tertiary hospitals and community health service centers in Guangzhou, China. Semi-structured interviews were conducted between January and February 2026. Data were analyzed using qualitative content analysis, supported by NVivo 14 software. Three descriptive categories were identified: (1) Contextual appropriateness of AI-simulated compassion: judgments of compassion and response suitability; (2) Transferability of emotional support: differing perspectives on task allocation and professional presence; and (3) Risks and responsibilities in the clinical application of AI-simulated compassion: safety safeguards and professional accountability. Healthcare professionals’ anticipatory ethical judgments regarding AI-simulated compassion extended beyond its emotional support functions to encompass the contextual appropriateness of responses, professional care relationships, and risks and responsibilities arising from ongoing interactions. This study highlights the ethical tension between technologically mediated emotional support and relational care, providing empirical insights into the boundaries of AI-simulated compassion in clinical practice and its implications for compassion and AI ethics. Future research should incorporate patient perspectives and firsthand clinical experience to further examine these ethical considerations and explore approaches that integrate contextual appropriateness, professional involvement, and ongoing accountability safeguards.