Warm but Redundant? How Different Generations Evaluate Elderspeak in AI Chatbots for Health Misinformation Correction

This study examines how linguistic and social cues shape relational and persuasive outcomes in human–AI communication for correcting health misinformation. Drawing on Communication Accommodation Theory, the Stereotype Content Model, and the Computers Are Social Actors paradigm, we test how elderspeak, voice age, and voice gender affect perceived warmth, competence, and information redundancy, which in turn influence emotional connection and misinformation correction. Two online experiments in China (N = 473 for older adults; N = 914 for general population) employed a 2 × 2 × 2 between-subjects design. Results show that elderspeak and feminine or middle-aged voices boost warmth and emotional connection but also increase redundancy, weakening correction outcomes. Mediation analyses are underway to clarify mechanisms. The findings advance theory by framing accommodation as a design variable and inform the creation of culturally sensitive chatbots for health communication.