Investigating the Impact of Social Media Authenticity and Sustainability on Perceived Value and Purchase Intention of International Brand Consumers Mediated by Trust

Abstract

To examine relationships between perceived social media message characteristics and consumer trust, perceived value, satisfaction, and purchase intention in the cosmetics and personal care context. A cross-sectional online survey included 384 respondents recruited through relevant social media channels and groups. Recruitment was based on online convenience access. Responses were collected using five-point Likert items and analyzed in SPSS and SmartPLS. Seven constructs were represented in the reported measurement model. Structural inference used bootstrapping with 5,000 resamples. Interpretation distinguishes individual structural paths from the originally proposed indirect effects. Cronbach’s alpha ranged from 0.895 to 0.927, composite reliability from 0.927 to 0.945, and average variance extracted from 0.722 to 0.774. The largest reported heterotrait–monotrait ratio was 0.735. The structural table reported t-statistics of 24.363 for trust–perceived value, 3.148 for trust–satisfaction, and 12.317 for trust–purchase intention. However, standardized coefficients, confidence intervals, and indirect-effect estimates were not reported. Two predictor–trust statistics differed between the original table and diagrams. The sustainability-related measure also combined descriptions of responsible communication with continuity and responsiveness. The reported results identify trust as a relevant construct for further investigation of consumer evaluations of social media communications. They do not establish the magnitude or direction of effects, mediation, or causal changes in purchasing. Interpretation is limited by nonprobability sampling, incomplete measurement documentation, and inconsistent structural-output reporting.

Keywords: social media; message authenticity; credibility; communication sustainability; consumer trust; perceived value; satisfaction; purchase intention.

Introduction

Social media allows brands and consumers to exchange product information, experiences, and evaluations. These interactions can influence how users interpret a brand’s communications, but engagement with a platform is not equivalent to trust or purchasing. Research on online brand communities distinguishes engagement with the brand from engagement with other users, emphasizing the need to identify the object of a consumer’s evaluation (Dessart et al., 2015). Research on influencer communication also links message value and source credibility to trust in branded content (Lou & Yuan, 2019). In the beauty and fashion context, Sokolova and Kefi (2020) examined credibility and parasocial interaction in relation to purchase intentions, providing a relevant comparison for social media research.

For cosmetics and personal care products, online messages may address ingredients, use, product performance, and consumer experience. The credibility of such information is relevant when users cannot directly evaluate a product before purchase. Online credibility assessment depends on how users evaluate information and its sources; visually appealing presentation alone does not establish reliability (Metzger, 2007). This provides a rationale for examining message characteristics in relation to trust without assuming that all consumers respond in the same way.

Perceived authenticity concerns whether consumers regard a brand or its communications as genuine. Morhart et al. (2015) conceptualized brand authenticity through credibility, integrity, symbolism, and continuity. Their framework also illustrates a measurement challenge for studies that treat authenticity and credibility as separate constructs. The distinction must be supported by item content and empirical evidence, rather than by different labels alone. The present study considers these constructs separately as represented in its reported measurement model; it does not claim to have administered the Morhart et al. scale.

Sustainability requires additional conceptual care. Environmental and ethical responsibility concerns a brand’s practices and claims, whereas communication continuity concerns regularity, consistency, and responsiveness over time. The study documentation uses the sustainability label alongside descriptions of sustained communication, dissemination of responsibility-related information, and responsiveness. These descriptions do not establish a distinct measure of environmental performance. Accordingly, sustainability is retained here as the original construct label, with interpretation restricted to the reported communication-related content. Environmental sustainability communication requires a separate conceptual basis: greenwashing concerns the mismatch between environmental communication and performance (Delmas & Burbano, 2011). Chen and Chang (2013) examined greenwashing in relation to green trust, consumer confusion, and perceived risk; these environmental constructs should not be equated with posting continuity.

Trust offers a theoretical connection between message evaluation and broader brand relationships. Morgan and Hunt (1994) positioned trust within relationship marketing, although their original empirical setting differs from consumer cosmetics purchasing. In the present context, authenticity and credibility may provide grounds for confidence in a brand, while trust may be related to perceived value and satisfaction. These relationships remain empirical questions rather than automatic consequences of digital exposure.

Perceived value refers to an evaluation of benefits relative to costs, while satisfaction reflects an evaluative response to experience or expectations. Purchase intention concerns a stated willingness to buy and must be distinguished from observed purchasing. The theory of planned behavior recognizes intention as related to, but distinct from, behavior (Ajzen, 1991). This study does not estimate that theory’s complete model because attitudes, subjective norms, and perceived behavioral control are not separately represented. The benefits–sacrifices interpretation of value follows Zeithaml (1988), while Sweeney and Soutar (2001) distinguish emotional, social, quality/performance, and price-related dimensions. Oliver (1980) provides an expectations–disconfirmation account of satisfaction. These frameworks clarify the concepts without implying that their original scales were administered in this study.

The study examines a proposed network of relationships among communication perceptions, trust, and consumer evaluations in an Iranian cosmetics and personal care context. Its contribution is a contextual examination of these relationships rather than a claim to the first comprehensive model in the field. Because recruitment did not establish an exclusively international-brand sample, the conclusions concern the recruited respondents and are not restricted to verified consumers of international brands.

Conceptual framework and scope of hypotheses

The conceptual rationale combines information signaling with relationship marketing. Signaling theory concerns communication under information asymmetry (Spence, 1973), while relationship marketing provides a basis for considering trust in ongoing exchange (Morgan & Hunt, 1994). Here, perceived message characteristics are positioned as antecedents of trust in the specified model, and trust is linked to perceived value, satisfaction, and purchase intention. The arrows represent proposed structural relationships; the cross-sectional design does not establish temporal ordering.

The original hypothesis set proposed that each of four predictors—authenticity, credibility, sustainability, and advertising effectiveness—was related through trust to each of three outcomes: perceived value, satisfaction, and purchase intention. This produces 12 proposed indirect relationships. The reported model, however, contains only the first three predictors. Advertising effectiveness is not represented in its measurement or structural results, and its three proposed indirect relationships are therefore outside the empirical coverage of the reported analysis.

The nine indirect relationships involving authenticity, credibility, and sustainability cannot be evaluated from individual-path statistics alone because specific indirect effects and bootstrap confidence intervals are unavailable. Perceived value is specified as an endogenous construct linked to satisfaction. No moderating interaction was estimated.

The structural specification comprises eight paths: authenticity to trust; credibility to trust; sustainability to trust; trust to perceived value; trust to satisfaction; trust to purchase intention; perceived value to satisfaction; and satisfaction to purchase intention. This specification also implies possible serial pathways through value and satisfaction, but those pathways are not empirically quantified in the reported results.

Methods

Design and recruitment

This applied study used a cross-sectional, descriptive-correlational survey. The target group comprised social media users, particularly Instagram users, with a history of interacting with brand pages or purchasing cosmetics and personal care products. Electronic questionnaires were distributed in relevant online channels and groups. This recruitment procedure is classified as online convenience sampling; a probability sampling frame and random-selection mechanism were not documented.

The reported sample comprised 384 respondents. This target was based on a general sample-size table for a large population, rather than a documented power analysis for the structural model. It therefore does not, by itself, demonstrate adequate power for the indirect relationships. Recruitment dates, detailed geographical coverage, invitation counts, response rates, exclusions, missing-data handling, and duplicate-response screening were not reported. The recruitment criterion did not require all respondents to have purchased an international brand.

Questionnaire and construct specification

The questionnaire included demographic questions on gender, age, education, and daily social media use, together with construct-related items rated on a five-point Likert scale from 1 (strongly disagree) to 5 (strongly agree). The reported measurement results cover seven constructs: social media authenticity, credibility, social media sustainability, consumer trust, perceived value, satisfaction, and purchase intention. Advertising effectiveness appeared in the initial conceptual description but was absent from the reported measurement results.

Authenticity was described through truthfulness, consistency between advertising and experience, and avoidance of exaggeration. The sustainability-related content was described through ongoing communication, dissemination of information about responsibility, and regular responsiveness. These descriptions support a cautious communication-based interpretation; they do not resolve the distinction between environmental responsibility and message continuity. No verified full item inventory, scale-adaptation record, translation procedure, or pilot-testing results accompany the reported analysis.

The measurement diagram used item labels q1–q33, with q33 appearing under both satisfaction and purchase intention. This duplication leaves the item-to-construct mapping uncertain. Figure 1 presents the structural specification without item labels; it does not resolve the measurement-specification uncertainty.

Statistical analysis

Descriptive analyses were conducted in SPSS and partial least squares structural equation modeling was conducted in SmartPLS. The reported constructs were specified as reflective. Internal consistency was summarized using Cronbach’s alpha and composite reliability; convergent validity was summarized using average variance extracted (AVE). Discriminant validity was reported using the heterotrait–monotrait ratio (HTMT). The present interpretation follows the distinction between measurement assessment and structural assessment described by Hair et al. (2022) and Henseler et al. (2015). Reporting guidance also distinguishes measurement assessment, structural inference, and prediction as separate evaluation tasks (Hair et al., 2019).

The reported structural analysis used bootstrapping with 5,000 resamples and a nominal 5% significance level. The original report referred to a two-sided t-statistic threshold of 1.96. Software versions, bootstrap interval method, random seed, and other estimation settings were not specified. Exact bootstrap p-values and confidence intervals were not supplied; values originally rounded to 0.00 are not presented as exact zero probabilities.

The results below preserve the reported aggregate measurement statistics and distinguish them from incomplete structural evidence. No standardized path coefficient, standard error, indirect effect, confidence interval, R², effect size, or predictive statistic has been reconstructed from t-values. A t-statistic concerns an estimate relative to its uncertainty and cannot establish its direction or substantive magnitude when the underlying coefficient is absent.

Assessment of indirect relationships

The hypothesized mediation concerns products of component paths rather than the significance of those paths considered separately. Specific indirect effects and their bootstrap confidence intervals are required to evaluate the proposed trust pathways, including any serial pathways involving perceived value and satisfaction (Hair et al., 2022). Those results were not reported. The analysis is therefore presented as an incomplete assessment of the proposed mediation framework, not as confirmation of the original 12 hypotheses. This distinction is consistent with contemporary mediation guidance for PLS path models (Nitzl et al., 2016). Moreover, cross-sectional mediation estimates may misrepresent longitudinal processes, so even a statistically supported indirect association would not establish a temporal mechanism (Maxwell & Cole, 2007).

Results

Respondent characteristics

The sample included 221 women (57.55%) and 163 men (42.45%). These percentages describe the recruited respondents and do not establish representativeness of the cosmetics market.

Respondents younger than 25 years accounted for 37.50% of the sample, and those aged 25–35 years accounted for 32.55%. Together, these categories represented 70.05% of respondents.

Bachelor’s degree holders comprised 185 respondents (48.18%). Reported proportions for postgraduate education, associate degrees, and diploma-level education or below were 17.70%, 5.72%, and 28.38%, respectively. Counts for these remaining categories were unavailable, preventing reconciliation of the small discrepancy in the percentage total.

A total of 228 respondents reported one to three hours of daily social media use (59.4%), and 20.1% reported more than three hours. These distributions describe use within the sample rather than population-level platform penetration or psychological engagement.

An additional descriptive statement reported a 50% preference for international brands, but a corresponding item-level breakdown was not provided. Brand preference does not establish exclusive international-brand exposure or purchasing. The study does not report a uniform reference brand for all trust, value, satisfaction, and intention responses.

Measurement results

Table 1 presents the matrix described in the original analysis as HTMT. Its values range from 0.366 to 0.735; the largest value concerns perceived value and consumer trust. The reported point estimates are below 0.85 and are consistent with empirical distinctiveness under that convention. Bootstrap HTMT intervals were not provided, and these point estimates do not establish conceptual validity or resolve overlapping item meanings (Henseler et al., 2015).

PV = perceived value; AU = authenticity; TR = consumer trust; CR = credibility; SA = satisfaction; PI = purchase intention; SU = the originally labeled sustainability construct. Diagonal dashes and blank upper-triangle cells denote values not applicable or not repeated.

Cronbach’s alpha ranged from 0.895 to 0.927, composite reliability from 0.927 to 0.945, and AVE from 0.722 to 0.774 (Table 2). These values exceed the conventional minimum values of 0.70 for internal consistency and 0.50 for AVE. Favorable aggregate indices do not substitute for inspection of item wording and individual loadings (Hair et al., 2022).

AVE = average variance extracted. Across constructs, Cronbach’s alpha ranged from 0.895 to 0.927, composite reliability from 0.927 to 0.945, and AVE from 0.722 to 0.774.

An indicator’s standardized loading describes its association with the construct; the squared loading represents the proportion of explained indicator variance. Individual standardized loadings were unavailable, and the measurement diagram displayed t-statistics rather than loading estimates. Item-level reliability therefore cannot be evaluated from the aggregate results alone.

Structural results and reporting consistency

Table 3 reproduces the t-statistics in the original structural table and explicitly identifies conflicting values in the diagrams. The trust–perceived value path had the largest listed t-statistic (24.363), followed by trust–purchase intention (12.317). These comparisons concern test statistics, not comparative effect strength. Without standardized coefficients, neither the magnitude nor the direction of these relationships can be established.

The three numeric columns reproduce the available tabular and graphical outputs. Differences for authenticity → trust and sustainability → trust exceed differences in displayed precision and remain unresolved. Figure 1 shows the structural specification without numerical estimates. Standardized coefficients, standard errors, exact bootstrap p-values, and confidence intervals are unavailable; these t-statistics alone do not establish the direction or magnitude of effects.

All t-statistics in the original structural table exceeded its nominal threshold of 1.96. Nevertheless, the missing coefficients, incomplete bootstrap output, and conflicting diagram values limit interpretation of the statistical evidence. The model does not provide reported results for advertising effectiveness or a direct perceived value–purchase intention path. The latter relationship is represented only indirectly through satisfaction in the displayed specification.

Status of the original indirect hypotheses

The mediating role of trust remains unconfirmed because specific indirect effects and their uncertainty intervals are unavailable. Neither indirect-only nor complementary mediation can be established, and serial mediation through value or satisfaction cannot be quantified.

Discussion

This study considered trust within a network linking perceived social media communications to value, satisfaction, and purchase intention. The reported aggregate reliability and AVE values were favorable, and the structural table contained t-statistics above the original nominal cutoff. These observations are narrower than a demonstration that message characteristics improve trust or cause purchases. The lack of standardized coefficients and indirect-effect estimates prevents conclusions about effect magnitude, direction, and mediation.

The theoretical relevance of trust is consistent with relationship marketing, where trust concerns confidence in an exchange relationship (Morgan & Hunt, 1994). The present findings do not replicate the full commitment–trust model, however, and a large trust-related t-statistic does not establish that trust is the most influential construct. It may reflect the coefficient, its precision, or both. Comparing substantive importance requires the actual estimates and relevant effect-size information.

Authenticity and credibility warrant separate conceptual attention. Morhart et al. (2015) included credibility within a broader authenticity framework, whereas this study represents the two as distinct constructs. The reported HTMT values offer preliminary evidence of empirical separation, but their meaning depends on the exact items. If the items largely ask whether the same brand information is honest, believable, and trustworthy, differences between construct labels may overstate theoretical distinctiveness. The full instrument is needed to evaluate that possibility.

The sustainability-related construct is the most consequential interpretive limitation. Regular communication and responsiveness may convey reliability, while disclosure of environmental responsibility may convey ethical commitment. Neither demonstrates actual sustainability performance. Combining these ideas within one label can make its structural relationship difficult to interpret. Consequently, the results do not support recommendations about renewable ingredients, recyclable packaging, or animal-welfare practices as tested determinants of purchase intention.

Perceived value and satisfaction occupy different positions in the reported model. Value is linked to satisfaction, which is in turn linked to purchase intention. Trust also has separately specified paths to all three. These relationships suggest possible parallel and serial mechanisms, but the present output cannot establish which mechanism contributes most or whether any specific indirect effect differs from zero. A conceptual sequence is not evidence of temporal mediation.

The recruitment design further limits the scope of interpretation. Respondents were recruited through online channels rather than sampled from a defined consumer population. Their brand familiarity, prior purchases, income, product category, and reference brands could affect both message evaluations and intention. The observed demographic distribution does not remove selection bias, and preference for international brands does not validate an international-brand-specific model.

Practical implications

The framework identifies communication features that managers can consider when assessing their social media practices. Clear product descriptions, appropriately documented claims, transparent responses to questions, and consistency between promotional messages and user experience are reasonable areas for attention. Online credibility research provides a rationale for helping users evaluate information (Metzger, 2007), while brand-community research emphasizes interactions with both brands and other users (Dessart et al., 2015). These considerations are not presented as interventions proven effective by this survey.

Regular communication can be evaluated separately from environmental or ethical claims. A content calendar may help coordinate information, but posting frequency should not be treated as evidence of environmental sustainability. Similarly, user-generated content and direct responses may support dialogue without guaranteeing trust or purchase. The study does not compare these practices against discounts, paid campaigns, or alternative marketing approaches.

Evaluation of practical initiatives should distinguish exposure and interaction from consumer attitudes and actual behavior. Purchase intention is a self-reported outcome, and the study provides no transaction, repeat-purchase, or loyalty measures. The present results therefore do not establish improvements in sales, long-term customer retention, or the return on marketing expenditure.

Limitations

First, the cross-sectional design does not establish temporal order or causation. Prior satisfaction or familiarity may influence perceived authenticity and trust. Second, collecting all constructs through one self-report questionnaire creates a potential common method bias concern. Its presence and magnitude cannot be inferred from the design alone, and no documented assessment quantifies this risk (Podsakoff et al., 2003).

Third, online convenience recruitment limits generalizability. Recruitment timing, detailed coverage, exclusions, missing-data procedures, and duplicate screening were not reported, and satisfaction responses may not all refer to verified purchase experiences. Fourth, the complete item inventory and scale-development documentation are absent, the sustainability definition is mixed, and the repeated q33 label leaves the item mapping unresolved.

Fifth, the structural report lacks coefficients, uncertainty intervals, R², effect sizes, collinearity diagnostics, and predictive evaluation. R² concerns explained variance and is not a general measure of model fit; no claim of strong explanatory or predictive performance is warranted here. Two predictor–trust statistics remain inconsistent across outputs. Finally, the original mediation hypotheses were not accompanied by indirect-effect estimates, and advertising effectiveness was absent from the reported model. These limitations preclude treating the study as a complete test of the proposed mediation framework. Out-of-sample predictive performance requires a dedicated assessment, such as the procedures discussed by Shmueli et al. (2019), rather than an inference from statistically significant paths.

Further research can distinguish message continuity from environmental responsibility, use a documented questionnaire and an explicit reference brand, and evaluate alternative temporal orderings. Longitudinal or experimental designs and observed purchase outcomes would strengthen assessment of the proposed mechanisms. Complete reporting of the measurement and structural outputs is necessary before the model can support effect-size comparisons or mediation claims.

Conclusion

The study reports aggregate measurement statistics and selected structural-path test statistics for seven constructs among 384 online respondents in the cosmetics and personal care context. Trust is central to the specified model, but its mediating role has not been demonstrated by reported indirect-effect tests. The findings should be interpreted as a limited assessment of proposed consumer-perception relationships, with unresolved measurement and structural-reporting issues, rather than evidence that social media authenticity or sustainability causes purchases.

Research transparency

Ethics and participant protection

Participation was voluntary, and informed consent was obtained before questionnaire completion. Participants’ confidentiality was protected, and responses were used solely for research purposes.

Funding and competing interests

This research received no external funding.

Data availability

Aggregate results are presented in the article. Access to participant-level data is subject to participant confidentiality and applicable consent restrictions.

Use of artificial intelligence

OpenAI ChatGPT was used for English-language editing, organization, reference and consistency checks, and preparation of the structural diagram. It was not used to generate participant data, statistical estimates, or ethics documentation, or to reanalyze the survey data.