Virtual reality’s impact on visit intention and the moderating role of social signal

This study examines how virtual reality (VR) tourism experiences are associated with visit intention by integrating perceptual stimuli, differentiated organismic mechanisms, and social-contextual moderation within an extended stimulus-organism-response framework in Vietnam. Using standardized pre-survey VR exposure followed by a survey, data were collected from 699 Vietnamese participants and analyzed using partial least squares structural equation modeling (PLS-SEM). The model incorporates vividness and interactivity as stimuli, presence, perceived authenticity, emotional involvement, and enjoyment as parallel organismic states, and social signal as a moderating factor. The results reveal an asymmetric pattern of relationships within the specified parallel model. At the stimulus side, interactivity showed a larger indirect association with visit intention through the organismic layer (beta = 0.489) than vividness (beta = 0.346). These coefficients refer to indirect associations within the specified S–O-R model, not to direct effects or explained variance. Within this specification, perceived authenticity has the largest organismic coefficient (beta = 0.502), followed by presence (beta = 0.264), emotional involvement (beta = 0.219), and enjoyment (beta = 0.182); the four interaction terms display positive signs for authenticity and emotional involvement and negative signs for presence and enjoyment. The model explains 57.9% of the variance in visit intention. However, estimated-model fit is weak (SRMR = 0.134; NFI = 0.768); all coefficient magnitudes, rankings, indirect associations, and interaction signs are therefore interpreted as exploratory estimates conditional on this potentially misspecified model, not as stable or uniquely identified structural effects. The study consequently offers a cautious account of concurrent organismic pathways rather than claiming a validated structural hierarchy. The lower-order social-signal term was negative (β = − 0.309, p < 0.001, f2 = 0.090) within the same specification; no model-invariant social-signal effect is claimed.