Dialect or Mandarin: virtual influencers' language fit with tourist destination type
Two scenario-based experiments (240 then 480 participants) show that the effect of a virtual influencer's language on advertising effectiveness depends on destination type, an interaction statistically associated with perceived spatial presence and asymmetrically moderated by influencer gender.
Results
In both studies, the interaction between language (dialect vs. Mandarin) and destination type (cultural vs. natural) is significant. Study 1 used two destinations in Sichuan as the experimental setting, Chengdu Taikoo Li as the cultural destination and Jiuzhaigou as the natural destination, with a single female virtual influencer whose gender was held constant across the four conditions. Average advertising effectiveness there is higher with dialect (M = 4.864) than Mandarin (M = 3.903) for the cultural destination, and higher with Mandarin (M = 5.189) than dialect (M = 4.314) for the natural destination.
The authors report a statistical mediation by perceived spatial presence: in Study 1, the index of moderated mediation is significant. Spatial presence is measured, not manipulated, and the language × destination interaction is randomized; this result establishes a statistical association reported by the authors, not a demonstrated causal mechanism.
In Study 2 (Hong Kong Palace Museum as the cultural destination, Hong Kong Wetland Park as the natural destination, 480 participants), the same pattern recurs, averaged across both influencer genders: dialect outperforms Mandarin for the cultural destination (M = 4.922 vs. 4.228), while Mandarin outperforms dialect for the natural destination (M = 4.819 vs. 4.101).
Study 2 further reveals a significant three-way interaction between language, destination and influencer gender: for the cultural destination, the dialect advantage is significantly stronger for female than male influencers, who show no significant difference between dialect and Mandarin. For the natural destination, Mandarin outperforms dialect for both genders, with no significant difference between them: the gender-moderation hypothesis is not supported there.
Critical reading
Each study uses a single destination per category and an exclusively Chinese sample, so the destination-type effect remains confounded with the specific sites and this sample.
The authors measured perceived comprehension difficulty and self-reported dialect familiarity, not objective comprehension or regional identity, and did not control pitch, speaking rate, vocal naturalness or dialect intensity.
In Study 2, each influencer gender is embodied by a single avatar whose appearance and voice change together: the observed gender effect reflects overall gender presentation, not the separate effect of a visual or vocal cue.
Neither experiment included a human-endorser condition, and authenticity, warmth, competence, psychological proximity and processing fluency were not measured, which prevents establishing that the effect is unique to virtual influencers.
The advertisement videos used as stimuli are not included in the text, only referenced in an appendix: their exact content remains unverified by this analysis.
Application
For a culturally dominant destination, the authors suggest having the virtual influencer speak in regional dialect, to strengthen the match between language and destination attributes.
For a naturally dominant destination, they suggest Mandarin, easier to understand for audiences with varied linguistic backgrounds. They warn that the cultural-destination gender asymmetry is the reliable finding, that the male advantage in natural destinations should not be over-generalized, and recommend weighing influencer gender together with language type, destination characteristics and advertising objectives rather than adopting a fixed gender-based preference.
Research & evidence
The study analysed
Speaking like a local: how virtual influencers’ dialect use shapes advertising effectiveness
Read the original study Full textContinue reading
All analyses →
