AI virtual digital human influencers vs. human influencers: the impact of health short videos on older adults' cognition and attitude.

Introduction: This study examines differences between AI Virtual Digital Human (AI-VDH) influencers and physician influencers in terms of recognizability, credibility assessment, and their influence on health cognition and attitudes among older adults in Chinese short-video contexts. Methods: A mixed-methods approach combining questionnaires ( Results: Older adults generally expressed confidence in distinguishing AI-generated figures from real persons; however, they often struggled to establish stable recognition criteria in highly realistic scenarios and relied on platform prompts to calibrate their judgments. Credibility assessments were influenced by both content accessibility and life experience, while medical credentials and institutional endorsements functioned as important authoritative cues. In AI-VDH scenarios, participants were more sensitive to presentation cues such as voice, facial expressions, and movements, which prompted greater caution. Both video types produced high self-reported improvements in health knowledge and attitudes, although physician-presented videos achieved slightly higher evaluations. Discussion: This study extends the dual-processing-three-level framework to embodied AI health video contexts by emphasizing the distinction between subjective confidence and actual recognition performance, while incorporating platform labeling into credibility calibration mechanisms. The findings reveal a calibration bias characterized by "high confidence but unstable recognition standards" among older adults in AI-VDH contexts. Platform prompts such as "suspected AI generation" play an important role in credibility calibration. The study provides practical implications for optimizing platform labeling systems, standardizing health science communication content, and developing AI and health literacy interventions for older adults.