A 16-item unidimensional scale measuring consumer perceptions of the trustworthiness of virtual social influencers (VSIs) -- computer-generated or digitally altered digital personas used in brand endorsement and social media marketing. Developed using an algorithmic item generation approach (the Psychometric Item Generator, PIG) seeded with GPT-4-generated items and refined through expert panel validation, EFA (n = 400), CFA (n = 297), nomological validity testing (n = 306), and pre-registered predictive validity testing (n = 210). Trustworthiness is conceptualised as a cognitive assessment of a VSI's credibility, reliability, and integrity, distinguishing it from the affective construct of connectedness (see companion scale VSIC, ID 1892). Grounded in socio-technical systems theory, which frames VSI trustworthiness as shaped by the interplay of consistent messaging, transparent disclosure, and culturally resonant content. In nomological testing, trustworthiness significantly predicted authenticity perceptions (beta = .512) and truthful endorsement (beta = .742). In predictive validity testing, it was positively associated with intention to interact (beta = .404), but did not significantly predict recommendation intention, advice-following intention, or overall behavioral intentions -- suggesting trust establishes initial credibility but that emotional bonding (captured by VSIC) drives most downstream behavioral outcomes.
Likert scale (exact point range not specified in the focal paper; mean item scores range from 4.26 to 4.74, consistent with a 7-point format)
Full scale items shown. Always cite the original source when using this scale in research.
The VSIT and its companion connectedness scale (VSIC, ID 1892) together constitute the full Virtual Social Influencer (VSI) scale introduced by van Esch et al. (2026). The two factors are theoretically distinct (cognitive vs. affective) but empirically correlated (r = .878 in CFA). HTMT ratio = .88, below the conservative .90 threshold, confirming discriminant validity. Items were generated algorithmically via the Psychometric Item Generator (PIG) -- an NLP-based tool adapted from the psychology literature and introduced here into marketing scale development -- with GPT-4-generated seed items and expert panel refinement (n = 20, 45% female, 90% aged 18-40). Note that TR09 ('I feel a sense of satisfaction or pride in following VSIs') loads on the trustworthiness factor despite its affective surface content; this reflects the broad credibility-and-regard conceptualisation used by the authors. Samples collected via Credamo (Studies 2, 3, 5) and CloudResearch (Study 4), predominantly Chinese, aged 18-40; cross-cultural validation recommended before applying in Western or older demographic contexts.
van Esch, P., Cui, Y., Arli, D., & Eskridge, B. (2026). Virtual social influencer scale development and validation: An algorithmic approach. Australasian Marketing Journal. Advance online publication. https://doi.org/10.1177/14413582261423028
Free access to 2455+ psychometric scales at PsychScales