Older adults’ social responses to AI virtual humans in digital kinship contexts: associations of loneliness, perceived cues and appraisals with purchase intention

As AI virtual humans become increasingly prominent communicative actors on short-video platforms, digital kinship content characterized by kinship-based forms of address, family narratives, caring expressions, and companionship-oriented interaction has begun to enter older adults’ everyday media use and become intertwined with commercial recommendation. Selectively drawing on the three-factor theory of anthropomorphism and integrating it with the Computers Are Social Actors (CASA) paradigm, this study conceptualizes loneliness as a manifestation of sociality motivation; emotional appeal, human likeness, and homophily as perceived cues and appraisals associated with elicited agent knowledge; and social presence and empathy as proximal social responses within the CASA framework. It examines older adults’ social perceptions of and emotional responses to kinship-oriented AI virtual humans and their associations with purchase intention. Using a standardized set of AI virtual human short videos representing key features of digital kinship content as stimuli, we collected valid data from 372 Chinese adults aged 65 years and older and employed partial least squares structural equation modeling (PLS-SEM) and fuzzy-set qualitative comparative analysis (fsQCA) as parallel and complementary analytical approaches. The PLS-SEM results revealed distinct antecedent patterns for the two proximal responses. Emotional appeal showed the strongest association with social presence, whereas homophily was more strongly associated with empathy than with social presence. Human likeness was significantly associated only with social presence, while loneliness was positively associated with both responses. Empathy showed a stronger association with purchase intention than social presence, and all indirect associations were significant except the pathway from human likeness through empathy. The fsQCA results further showed that the combination of emotional appeal and empathy constituted the most stable configuration associated with high purchase intention, whereas loneliness emerged as a core condition only when homophily was also present. This study provides empirical evidence for understanding perceptions of social presence and empathy in later-life human–AI relational perception and offers practical implications for the governance of kinship-oriented AI commercial content and the protection of older consumers.