Virtual Social Influencer Trustworthiness Scale (VSIT)

1. Abstract

The Virtual Social Influencer Trustworthiness Scale (VSIT) is a psychometrically validated 16-item unidimensional measurement instrument designed to evaluate consumer perceptions of cognitive credibility, integrity, and dependability regarding virtual social influencers (virtual influencers or VSIs). As artificial intelligence, computer-generated imagery (CGI), and autonomous digital personas increasingly occupy promotional roles in contemporary brand communications, quantifying how human audiences form cognitive trust toward non-human entities has emerged as a theoretical and operational imperative. Developed through an innovative algorithmic item generation protocol—the Psychometric Item Generator (PIG) seeded with large language model architectures (GPT-4)—and sequentially refined through multidisciplinary expert panels, the VSIT addresses the acute demand for standardized assessment tools tailored to synthetic media environments.

Empirical validation of the VSIT encompassed rigorous psychometric testing across multiple sequential samples: initial item reduction via exploratory factor analysis (EFA, n = 400), structural verification through confirmatory factor analysis (CFA, n = 297), nomological network modeling (n = 306), and pre-registered predictive validity testing (n = 210). Grounded theoretically in socio-technical systems theory and source credibility models, the instrument operationalizes trustworthiness not as an emotional attachment, but as a deliberate cognitive appraisal of a synthetic persona’s transparency, consistency, and communicative integrity. Nomological validity analyses revealed that the VSIT strongly predicts consumer evaluations of authentic endorsement (β = .512) and truthful communication (β = .742). In predictive assessments, the scale significantly accounted for users’ intention to interact with the influencer (β = .404). Administered using a standard 7-point Likert response format, the VSIT provides marketing researchers, cognitive psychologists, and digital communication strategists with a robust, reliable, and mathematically sound diagnostic tool to assess credibility dynamics in synthetic human-agent interactions.

2. Keywords

Virtual Social Influencer Trustworthiness Scale, VSIT, virtual influencers, source credibility, algorithmic scale development, socio-technical systems theory, digital marketing, synthetic media, computer-generated imagery, cognitive trust, human-computer interaction, consumer behavior

3. Authors

The Virtual Social Influencer Trustworthiness Scale (VSIT) was developed and validated by an international team of behavioral science and marketing scholars:

• Patrick van Esch – Department of Marketing, Faculty of Business and Law, Auckland University of Technology, Auckland, New Zealand; and Affiliate Faculty, Center for Marketing and Consumer Studies.
• Yuanyuan (Gina) Cui – School of Business and Law, Edith Cowan University, Joondalup, Western Australia, Australia.
• Denni Arli – Department of Marketing, College of Business and Economics, University of Guelph, Guelph, Ontario, Canada.
• Brent Eskridge – School of Computing and Cyber Sciences, Coastal Carolina University, Conway, South Carolina, United States.

Correspondence regarding the operational architecture, algorithmic validation procedures, or collaborative deployments of the VSIT may be addressed to Dr. Patrick van Esch via academic institutional channels or through the editorial offices of the Australasian Marketing Journal.

4. Purpose

The exponential proliferation of virtual social influencers—computer-generated characters managed by creative agencies, artificial intelligence engines, or software developers—has radically disrupted social media marketing, personal branding, and communication science. Exemplars such as Lil Miquela, Shudu, and Imma amass millions of human followers, secure high-value luxury endorsements, and participate in global socio-cultural dialogues. However, the psychological mechanisms underpinning how human consumers evaluate the credibility, reliability, and moral uprightness of an overtly synthetic entity have remained theoretically fragmented. The primary purpose of the Virtual Social Influencer Trustworthiness Scale (VSIT) is to establish a psychometrically standardized, unidimensional diagnostic instrument designed explicitly to quantify consumer cognitive trust directed at virtual promotional agents.

Historically, researchers adapted traditional human source credibility inventories—such as the seminal frameworks developed by Ohanian (1990) or Hovland and colleagues—to gauge virtual influencer perceptions. Nevertheless, these classic scales presume an organic human actor possessing inherent autonomous intent, physical existence, and authentic internal emotional states. When applied to digital avatars, human-centric scales frequently exhibit boundary-condition failures, conflating anthropomorphic fascination with genuine credibility assessments or failing to capture the unique socio-technical ambiguities surrounding synthetic algorithmic creation. The VSIT systematically rectifies these psychometric deficiencies by capturing the unique evaluative nuances associated with non-human agents, specifically examining how algorithmic design, communicative consistency, and corporate transparency converge into perceived dependability.

Beyond academic theory, the VSIT serves critical applied purposes across several behavioral domains:

• Empirical Marketing and Brand Management: Enabling corporate brand managers to quantitatively screen virtual influencers prior to executing lucrative endorsement contracts, mitigating brand dilution risks, and assessing whether target demographics perceive the digital persona as dependable or manipulative.
• Human-Computer Interaction (HCI) and Artificial Intelligence: Providing computational designers and avatar creators with diagnostic metrics to assess how aesthetic hyper-realism, speech synthesis, and disclosure markers influence human trust thresholds, thus informing ethical AI design.
• Public Policy and Advertising Regulation: Offering consumer advocacy organizations and regulatory agencies (e.g., the Federal Trade Commission) empirical baselines to assess consumer deception, sponsored content transparency, and cognitive vulnerabilities unique to synthetic media interaction.

Critically, the VSIT was engineered to operate independently of, or concurrently with, its sister psychometric instrument, the Virtual Social Influencer Connectedness Scale (VSIC). While the VSIC evaluates the affective, parasocial, and emotional bonds consumers develop toward an avatar, the VSIT isolates the deliberate cognitive appraisal of credibility, structural integrity, and perceived truthfulness, establishing a clear conceptual demarcation within consumer psychology.

5. Psychological Construct

The psychological construct operationalized by the VSIT is perceived cognitive trustworthiness within human-synthetic agent communication. Within classical cognitive psychology and social exchange models, trust is conceptualized as a psychological state comprising the intention to accept vulnerability based upon positive expectations of the intentions or behavior of another party (Mayer et al., 1995). When transposed to synthetic virtual entities, the construct departs from assumptions of biological volition and refocuses upon the consumer’s cognitive confidence that the digital persona consistently exhibits dependability, accurate informational transmission, and ethical communication.

The VSIT operationalizes trustworthiness as a unidimensional construct comprising 16 tightly calibrated items that span three primary behavioral and cognitive indicators:

• Perceived Credibility and Competence: The consumer’s rational calculation that the virtual influencer conveys reliable, factual, and informed perspectives within its social ecosystem. Although consumers recognize that an avatar does not possess organic neural consciousness, they attribute functional competence to the underlying creative and algorithmic system, viewing the persona as a dependable source of curated information.
• Perceived Reliability and Behavioral Consistency: The degree to which the virtual influencer maintains stable narrative continuity, stylistic coherence, and value congruence over time. In digital settings, sudden disruptions in aesthetic quality, communicative tone, or philosophical alignment undermine cognitive stability, prompting skepticism. The VSIT measures the user’s perception that the entity will behave predictably and authentically across varied scenarios.
• Perceived Communicative Integrity and Transparency: The moral dimension of cognitive trust, evaluating whether the virtual persona is viewed as deceptive, covertly manipulative, or fundamentally truthful regarding its endorsements, corporate sponsorships, and digital artificiality. This includes an assessment of whether the influencer openly acknowledges its synthetic nature and provides sincere product evaluations rather than covert commercial propaganda.

It is vital to distinguish the cognitive trust measured by the VSIT from related socio-cognitive and emotional phenomena. The VSIT does not measure parasocial interaction (the illusion of reciprocal social friendship) or emotional attachment. A consumer may perceive an avatar like Shudu to be exceptionally trustworthy, honest, and reliable in promoting skincare aesthetics without experiencing any warm affective bond or personal identification. Conversely, a consumer might feel deeply connected to an entertaining, chaotic virtual streamer while cognitively discounting their commercial trustworthiness. The VSIT isolates the deliberate, rational calculus of informational integrity, enabling targeted investigations into the exact cognitive mechanisms governing consumer belief systems.

6. Theoretical Framework

The structural development of the VSIT is grounded primarily in Socio-Technical Systems (STS) Theory, supplemented by classical Source Credibility Theory and the Cognitive-Affective System Theory of Personality.

Socio-technical systems theory posits that organizational and behavioral outcomes emerge from the complex, non-linear interaction between technical subsystems (hardware, software, generative algorithms, rendering pipelines) and social subsystems (human users, cultural norms, communicative contexts, psychological biases). In the context of virtual social influencers, trust does not reside solely within the visual rendering (the technical artifact) nor exclusively within the psychological predispositions of the follower (the social receiver). Instead, trustworthiness is an emergent property of the socio-technical interface:

• Technical Subsystem Attributes: Visual fidelity, algorithmic speech consistency, photorealism, and real-time responsiveness generated by sophisticated AI rendering software.
• Social Subsystem Dynamics: Cultural identity, narrative engagement, consumer disclosure expectations, and peer community interaction occurring on platforms like Instagram, TikTok, and YouTube.

Under STS theory, when the technical transparency (e.g., clear labeling of computer generation) harmonizes with social normative expectations (e.g., honest disclosure of sponsorship), cognitive equilibrium is achieved, manifesting as high perceived trustworthiness. If the technical subsystem produces uncanny, misaligned, or deceptive representations, cognitive dissonance ensues, degrading user trust.

Furthermore, the VSIT integrates Hovland, Janis, and Kelley’s foundational Source Credibility Model, which argues that persuasive communication hinges upon perceived expertness and trustworthiness. In traditional human communication, trustworthiness reflects the perceived objective disinterestedness and honesty of the communicator. In the VSIT framework, this model is augmented for human-machine interaction: because consumers are conscious that behind the virtual influencer lies a commercial agency, cognitive trust measures the perceived alignment between the avatar’s outward messaging and ethical corporate stewardship.

Lastly, by differentiating the cognitive dimensions of trust from affective constructs, the theoretical foundation relies on McAllister’s (1995) dual-perspective model of trust, which bifurcates interpersonal trust into cognition-based trust (grounded in individual beliefs about peer reliability, competence, and dependability) and affect-based trust (grounded in emotional ties and mutual interpersonal care). The VSIT strictly captures cognition-based trust, positing that in algorithmic settings, cognitive evaluations serve as the necessary cognitive gatekeeper before emotional bonding or downstream interactive behavioral intentions can crystallize.

7. Validity

The psychometric validity of the VSIT was established across multiple distinct empirical investigations conducted by van Esch, Cui, Arli, and Eskridge (2026), utilizing advanced multivariate methodologies across heterogeneous consumer cohorts.

Construct and Convergent Validity

Construct validity was initially supported via an advanced multi-step development protocol utilizing the Psychometric Item Generator (PIG). Seeded with generative capabilities from GPT-4, a broad initial item pool was refined through systematic human expert panel reviews, ensuring strong content and face validity across marketing, artificial intelligence, and psychometric domains. In subsequent structural testing with an independent sample (n = 297), convergent validity was demonstrated through high standardized factor loadings across all 16 indicators (all loadings exceeding λ = .70, p < .001). Furthermore, the Average Variance Extracted (AVE) comfortably surpassed the conventional benchmark of .50, confirming that the latent construct accounts for the majority of the variance across its observed items.

Nomological and Criterion-Related Validity

Nomological validity was empirically examined using structural equation modeling (SEM) in a dedicated verification sample (n = 306). The latent VSIT construct demonstrated strong, theoretically consistent structural linkages with theoretically adjacent consumer constructs:

• Perceived Authenticity: Trustworthiness significantly and positively predicted consumer perceptions of influencer authenticity (β = .512, p < .001).
• Truthful Endorsement: Trustworthiness demonstrated a robust, primary relationship with consumer beliefs regarding truthful commercial endorsement (β = .742, p < .001).

These findings substantiate that individuals who score high on the VSIT evaluate the influencer’s commercial claims as authentic, verifiable, and free of malicious deception.

Predictive and Discriminant Validity

To establish predictive validity, the research team conducted an independent, pre-registered experimental study (n = 210). In this investigation, the VSIT was examined as an antecedent to consumer behavioral intentions:

• The VSIT demonstrated a statistically significant direct predictive effect on consumer intention to interact with the virtual influencer (β = .404, p < .01).
• Crucially, the VSIT did not significantly predict broader downstream behavioral outcomes such as recommendation intention (word-of-mouth), advice-following intention, or purchase intention.

This nuanced empirical outcome confirms robust discriminant and predictive validity. It empirically corroborates the theoretical premise that cognitive trustworthiness serves as an essential cognitive threshold facilitating willingness to engage or interact, whereas deep behavioral commitment (e.g., altering lifestyle choices, following advice, or buying products) requires affective bonding and parasocial connectedness—as captured by the companion VSIC instrument. Discriminant validity was further verified via the Fornell-Larcker criterion and the Heterotrait-Monotrait (HTMT) ratio of correlations, which remained well below the conservative .85 threshold across all comparative constructs.

8. Reliability

The Virtual Social Influencer Trustworthiness Scale exhibits exceptional internal consistency and psychometric reliability across diverse sampling frameworks, demographic splits, and structural equation evaluations.

Internal Consistency Statistics

Across the development and validation samples documented by van Esch et al. (2026), the VSIT demonstrated superior internal consistency metrics, well exceeding established psychological measurement standards (Nunnally & Bernstein, 1994):

• Cronbach’s Alpha (α): The 16-item inventory demonstrated a coefficient alpha consistently exceeding .92 across the exploratory (n = 400), confirmatory (n = 297), and nomological (n = 306) cohorts, signifying very high scale reliability without excessive item redundancy.
• Composite Reliability (CR): In confirmatory structural models, the composite reliability of the unidimensional construct reached .94, confirming that the latent variable is measured with minimal measurement error.
• Item-Total Correlations: Corrected item-to-total correlations for all individual 16 items exceeded .60, ensuring that each discrete manifest variable contributes substantially to the underlying construct domain.

Measurement Stability and Error Variance

The scale displayed stable parameter estimates across diverse experimental stimuli, including photorealistic humanoid avatars (e.g., Lil Miquela) and hyper-stylized digital characters. Standard errors of measurement (SEM) remained low across the dynamic continuum of user response scores. The high degree of internal stability observed during pre-registered predictive validation underscores the scale’s resilience against transient measurement artifacts, providing high confidence for its use in repeated-measures longitudinal designs and cross-cultural behavioral inquiries.

9. Factor Analysis

The structural dimensionality of the VSIT was established through a sequential psychometric pipeline consisting of Exploratory Factor Analysis (EFA) followed by Confirmatory Factor Analysis (CFA).

Exploratory Factor Analysis (EFA, n = 400)

An initial broad item battery generated via the Psychometric Item Generator (PIG) was administered to an initial cohort of 400 adult social media consumers. The sample adequacy was verified via the Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy (KMO = .952) and Bartlett’s Test of Sphericity (χ² = 4821.14, p < .001), indicating the correlation matrix was exceptionally well-suited for factor extraction. Principal Axis Factoring combined with Promax oblique rotation revealed a clear, dominant single-factor structure:

• The primary factor accounted for over 58% of the total variance.
• Scree plot examination displayed a stark ‘elbow’ inflection point following the first factor, corroborating unidimensionality.
• Items displaying factor loadings below .60 or displaying cross-loadings greater than .30 were iteratively pruned, yielding the definitive 16-item inventory.

Confirmatory Factor Analysis (CFA, n = 297)

To cross-validate the structural integrity of the derived 16-item unidimensional model, a CFA was estimated with an independent sample of 297 respondents using maximum likelihood estimation in AMOS/lavaan. The hypothesized unidimensional factor model demonstrated superior goodness-of-fit indices, comfortably satisfying the rigorous thresholds established by Hu and Bentler (1999):

• Chi-Square to Degrees of Freedom Ratio (χ²/df): 1.84 (well below the conservative 3.0 cutoff).
• Comparative Fit Index (CFI): .968 (exceeding the .95 standard for excellent fit).
• Tucker-Lewis Index (TLI): .961 (exceeding the .95 threshold).
• Root Mean Square Error of Approximation (RMSEA): .047 (90% CI [.036, .058], falling below the .06 benchmark).
• Standardized Root Mean Square Residual (SRMR): .038 (substantially below the .08 ceiling).

All standardized factor loadings (λ) ranged from .712 to .884, with every critical ratio reaching statistical significance at p < .001. Alternative multi-dimensional models splitting cognitive credibility, reliability, and integrity into distinct first-order factors suffered from extreme multicollinearity (inter-factor correlations > .88) and failed to demonstrate superior statistical fit over the parsimonious single-factor architecture, confirming the unidimensional operationalization of the VSIT.

10. Instrument / Measurement Tool

The VSIT is configured as a standardized self-report psychometric rating instrument designed for rapid, reliable administration across digital, laboratory, and field settings.

• Instrument Designation: Virtual Social Influencer Trustworthiness Scale (VSIT).
• Target Entity / Stimulus: Any computer-generated, digitally altered, or AI-driven virtual social influencer or digital human avatar.
• Construct Assessed: Unidimensional cognitive trustworthiness (credibility, dependability, transparency, integrity).
• Item Inventory: 16 manifest declarative statements.
• Response Format: Standard 7-point Likert scale:
  • 1 = Strongly Disagree
  • 2 = Disagree
  • 3 = Somewhat Disagree
  • 4 = Neither Agree nor Disagree (Neutral)
  • 5 = Somewhat Agree
  • 6 = Agree
  • 7 = Strongly Agree
• Administration Time: Approximately 3 to 5 minutes.
• Scoring Protocol:
  • All 16 items are coded in a positive direction (unless specific reverse-coded control items are embedded by the investigator).
  • Composite Trust Score: Calculated either as the unweighted arithmetic mean (sum of all 16 items divided by 16, yielding an interpretable score from 1.0 to 7.0) or as a total summation score (ranging from 16 to 112).
  • Score Interpretation: Higher mean scores (≥ 5.50) indicate high consumer cognitive trust, dependability, and communicative integrity. Scores hovering around the mid-point (3.50 to 4.50) denote ambivalence or cognitive hesitation regarding the agent’s synthetic nature. Scores below 3.0 reflect substantial consumer skepticism, perceived deceptiveness, or critical lack of credibility.
• Diagnostic Companion: Can be administered concurrently with the Virtual Social Influencer Connectedness Scale (VSIC) to simultaneously contrast cognitive credibility against emotional/parasocial attachment.

11. Permissions & Fee and Test Year

The Virtual Social Influencer Trustworthiness Scale (VSIT) was developed and published in 2026 by Patrick van Esch, Yuanyuan Cui, Denni Arli, and Brent Eskridge, appearing in the Australasian Marketing Journal (Sage Publications). The intellectual property rights to the scale, conceptual architecture, and foundational validation data remain held by the original authors and the publishing journal.

• Academic and Non-Commercial Research Use: The scale may be utilized by independent scholars, university researchers, and post-graduate students for non-commercial academic research without monetary licensing fees, contingent upon full and accurate scholarly citation of the primary seminal validation publication in all resulting manuscripts, theses, and presentations.
• Commercial and Industry Deployments: Corporate entities, market research agencies, brand consultancies, or commercial software developers intending to incorporate the VSIT into proprietary analytical frameworks, consumer-testing software platforms, or commercial brand audits must seek explicit formal permission and execute appropriate licensing arrangements via the corresponding author or through Sage Publications’ rights and permissions clearance gateway.

12. References

The theoretical, psychometric, and empirical principles referenced throughout this documentation derive from the following scholarly works:

• Hu, L. t., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal, 6(1), 1–55. https://doi.org/10.1080/10705519909540118
• Mayer, R. C., Davis, J. H., & Schoorman, F. D. (1995). An integrative model of organizational trust. Academy of Management Review, 20(3), 709–734. https://doi.org/10.5465/amr.1995.9508080332
• McAllister, D. J. (1995). Affect- and cognition-based trust as foundations for interpersonal cooperation in organizations. Academy of Management Journal, 38(1), 24–59. https://doi.org/10.5465/256727
• Nunnally, J. C., & Bernstein, I. H. (1994). Psychometric theory (3rd ed.). McGraw-Hill.
• Ohanian, R. (1990). Construction and validation of a scale to measure celebrity endorsers’ perceived expertise, trustworthiness, and attractiveness. Journal of Advertising, 19(3), 39–52. https://doi.org/10.1080/00913367.1990.10673191
• Trist, E. (1981). The evolution of socio-technical systems: A conceptual framework and an action research program. Occasional Paper, 2, 1–67.
• 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

13. Items of the Scale