The role of social media in shaping purchasing behavior among consumers in the Kingdom of Bahrain

1 Introduction

The early twenty-first century has brought a fundamental shift in how people communicate and how economies operate (Kotler et al., 2021). Whereas earlier social media sites were mere connective tools between individuals, the current reality has seen the development of significant commercial social media sites which greatly impact conventional business dynamics (Abdalla et al., 2025; Wan et al., 2025). The convergence of peer interaction, content creation, and commercial transactions within these platforms has given rise to what is termed ’social commerce' – an environment where interaction, content, and commerce are in perfect harmony (Hajli, 2015; Liang and Turban, 2011).

It follows that the process would begin with need recognition through content recommendations and algorithms; information search would occur in branded pages and threads; alternative evaluation occurs in terms of social validation cues; purchase decisions occur within the context of social commerce sites; and finally, the post-purchase process involves review, comment, feedback, and electronic word-of-mouth (Cheung and Thadani, 2012; Kothari et al., 2025; Sokolova and Kefi, 2020; Zhang and Li, 2025).

Such challenges however take on greater importance in the case of Bahrain. It should be noted that the GCC country displays consistent high levels of social media penetration on a per-capita basis. Specifically, Bahrain holds a social media penetration of 97.4% ranking it fourth among countries globally as far as social media penetration is concerned (DataReportal, 2024), with social media sites such as Instagram, Snapchat, and Tiktok forming part of everyday life in the country. Nevertheless, little research has been done to examine the social media influence on consumer purchase decisions in Bahrain during various phases of the process. This is particularly important for the implications it poses to marketers in developing consumer content campaigns based on consumer data, the government in regulating the process, and academics in understanding the differences that exist between the GCC economies and Western ones. This research helps fill this gap by providing an empirical understanding of the influence of social media on consumer purchase behavior in Bahrain.

This study makes three distinct contributions to knowledge. Firstly, it provides an initial stage-differentiated analysis of social media influence on all five phases of consumer decision journey in Bahrain, helping to close the gap created by the absence of similar GCC literature on this matter. Secondly, whereas Mohammed (2025) studies age alone as the demographic variable affecting consumer decision making in Bahrain, the present study examines age, gender, income, and education simultaneously for the first time in Bahrain, creating the first multi-demographic convergence analysis of the GCC. Lastly, the study seeks to create a conceptual framework for demographic convergence, incorporating the algorithmic personalization and adding to existing theory in terms of the S-O-R model and ELM. The reason is simple; the exceptionally high social media penetration (97.4%) of Bahrain makes it an excellent candidate for such study (DataReportal, 2024).

There is an existing knowledge gap regarding the influence of social media on consumer behavior when it comes to making purchases and the particularities of the situation in the Bahraini market. The high reach of the social media platforms in no way guarantees that the reactions of the local consumers would be predictable as they could vary depending on the values, economic system, and digital competence of different generations. In addition to that, the growing prominence of the mechanisms of digital marketing creates uncertainty as to whether social media platforms are used more for thoughtful decision-making or impulsive spending.

• To determine the relationship between social media content exposure and purchase intention and impulse buying among consumers in Bahrain.
• To identify the social media platforms most frequently used and most influential in shaping Bahraini consumer purchase decisions.
• To examine differences in purchasing behavior attributable to gender, age, educational level, and household income.

2 Background and theoretical framework

2.1 Social media as a commercial ecosystem

Social media is defined as applications that operate via the internet where users create their own profiles, establish relationships, and communicate with others through engaging content (Kaplan and Haenlein, 2010). The transformation of social media into businesses was accompanied by a transition from old-school non-interactive broadcast advertisements towards interactive, algorithm-driven business-to-consumer communications (Kothari et al., 2025; Kotler et al., 2021). According to Zhang and Li (2025), social media ads play a direct role in the consumer's decision-making process concerning purchasing products by offering necessary information and emotional engagement. Social commerce can be described as a new type of business activities that emerge from the combination of social media communication and online shopping tools (Liang and Turban, 2011). In their paper, they described social commerce as a unique form of e-commerce that makes use of the power of social networking for purchase decisions. In his groundbreaking study, Hajli (2015) used the constructs of social commerce and revealed that all three factors – social support, ratings and reviews, and referrals had a significant effect on the intention to purchase, and social support played the most influential role among them.

2.2 Theoretical frameworks

2.2.1 Uses and gratifications theory (UGT)

Uses and Gratifications Theory (UGT) originally suggested in studies on mass communication implies that an audience consists of goal-oriented individuals who actively seek specific media forms for their needs in cognition, affect, and social interactions (Katz et al., 1974). Extending UGT to digital consumerism, prior descriptive accounts characterize Instagram as oriented toward aesthetic and hedonic gratifications, TikTok toward entertainment gratifications delivered through short-form video, and Snapchat toward immediacy and parasocial intimacy with influencers. These platform-gratification pairings are conceptual characterizations drawn from the broader UGT literature rather than classifications empirically tested within the present study, and they are treated here as background assumptions that motivate, but are not directly verified by, the analyses that follow.

In accordance with the findings reported by Wan et al. (2025), involvement in social media, which refers to making friends, self-exposure, and pleasure, has a positive impact on the intention of consumers to purchase something through self-disclosure (β = 0.168, p < 0.01). Different from the findings of other uses and gratifications studies, which were performed to investigate the association between making friends through television and print media use, the use of social media implies algorithmic mediation, which determines the kind of gratification available to consumers. Hence, the UGT was not designed to integrate mediation into motivations of consumers in terms of their gratification needs.

2.2.2 The stimulus–organism–response (S-O-R) model

The stimulus-oriented model presented by Mehrabian and Russell (1974) suggests that consumer behavior consists of three steps, including Stimulus (S), Organism (O) and Responses (R). In the context of social media, consumers' behavior is influenced by means of digital stimuli such as advertising, influencers, video trends, etc. and hence, their emotional processes (Abdalla et al., 2025; Kothari et al., 2025). For example, the analysis of 359 Saudi consumers, conducted by Abdalla et al. (2025), showed that social media trends, as well as the quality of information disseminated via social media, influence positively consumer behavior, which can be seen as an example of the influence of Stimulus (the use of trends in social media) on Response (consumer buying behavior). At the same time, according to Bui et al. (2025), the S-O-R model has been empirically proven in the field of visual electronic word-of-mouth communication and showed the positive impact of informativeness and entertainment of visual social media content (Stimulus) on online purchase intention (Response). As a result of structural equation modeling, it was found that Stimulus influences Responses more successfully if content is perceived visually than textually, which can be a reason why Instagram and TikTok have achieved great popularity within the current Bahraini environment. Therefore, it can be concluded that the extension of the S-O-R model described initially as an approach to influencing consumer behavior in retail stores (Mehrabian and Russell, 1974) should take into account not only the Stimulus but also its nature in social commerce environment. The major difference between Stimulus in retail and social commerce environments is the algorithmic selection and optimization of stimulus in the latter. In this regard, the extension of the model proposed can be considered extremely important in the case of Bahrain due to high platform penetration. It should be noted, however, that the present study operationalizes the S-O-R sequence only at its Stimulus (platform-specific content exposure) and Response (stage-level purchasing outcomes) endpoints; the Organism component – the internal cognitive and affective processing that theoretically mediates the two – is not directly measured and is instead inferred from the pattern of Stimulus-Response associations. This is acknowledged as a limitation of the present design in Section 7.

2.2.3 The elaboration likelihood model (ELM)

The Elaboration Likelihood Model (ELM), developed by Petty and Cacioppo (1986), describes two routes of attitudinal change and behavior influencing mechanisms: the central route is characterized by elaborate processing of the arguments contained within a particular message; while in the case of the peripheral route, contextual cues like attractiveness of the source, visual aesthetics, and elements of social proof are used to trigger attitudes without requiring any effort at all. The model can be easily applied to illustrate two main findings of the current research. First, ELM provides a theoretical basis for anticipating that platforms built around fast-paced, high-social-proof video content could generate purchase influence disproportionate to their usage share, since peripheral-route processing does not require sustained deliberate engagement; whether this pattern holds for TikTok in the present sample is examined empirically in Section 5.3. This phenomenon is explained by the activation of the peripheral route due to the fast pace of video content, the use of background sounds, and high social proof density, which leads to the generation of intentions to buy goods or services without requiring further analysis. Second, ELM offers grounds for predicting that education level would not moderate susceptibility to platform-driven persuasion: if social media usage is dominated by peripheral-route persuasion through commercially driven content, then individuals with higher education – whose resistance to persuasion typically operates through central-route counter-argument – would be expected to benefit comparatively little from that resistance. This prediction is tested as H3c and revisited in light of the empirical results in Section 6.

2.3 Integrated framework: UGT, S-O-R, and ELM in combination

In this current paper, UGT, S-O-R and ELM are combined in a tri-level model of explanation, which is as follows: (1) The use of UGT provides an opportunity to comprehend the motivational level, i.e., the motivations of consumers that explain why they are motivated to opt for some platform in pursuit of their cognitive, social and hedonic requirements; (2) The S-O-R approach describes the level of mechanism, explaining how, upon entrance into selected platform, commercial message leads to evaluation and arousal on an organic level, which, in its turn, triggers certain reactions on the part of consumers through purchase phases, and (3) finally, ELM explains the level of processing route, demonstrating the path by which a platform influences users (analytic/central vs. affective/peripheral), explaining why visual-intensive media platforms have overwhelming commercial efficiency and education does not moderate influence. Figure 1 depicts these three layers as they map onto the study's empirical design: the UGT layer corresponds broadly to RQ2 (platform selection and usage, tested via H2), the S-O-R layer corresponds to RQ1 (stimulus-to-response associations across the five purchasing stages, tested via H1) (Figure 2), and the ELM layer corresponds to RQ3 (processing-route differences reflected in the demographic moderation tests, H3). It should be emphasized that the empirical analyses reported in Section 5 test associations and between-group differences consistent with these three layers rather than directly measuring the internal mechanisms (motivational states, organismic processing, or route of elaboration) that the theories posit; the integrated framework is therefore used here as an interpretive lens for the observed patterns rather than as a fully operationalized, mediation-tested model, and this more modest positioning of the theoretical contribution is carried through to the Discussion and Limitations (Sections 6, 7).

2.4 Social media influence across the consumer decision journey

A growing body of research demonstrates the influence of social media platforms across all five phases of the consumer decision-making process (Engel et al., 1968), although to varying extents. Need recognition can now easily be triggered by algorithm-generated recommendations of images and videos. Information gathering is largely done online; according to numerous scholars, well over 75 percent of buyers browse social media websites prior to purchase (Zhang and Li, 2025). User ratings and eWOM communications act as trust signals to decrease the risk involved in the purchase (Kothari et al., 2025; Sokolova and Kefi, 2020). Cheung and Thadani (2012) provide an exhaustive review of existing studies on the effects of eWOM and found that the level of eWOM informativeness, credibility, and amount were three major drivers of behavioral change in consumers, while evaluation of alternatives was identified as the phase at which eWOM influence tends to peak – a pattern this study's stage-level analysis (Section 5.2) also examines. In George et al.'s (2025) systematic review on the success of influencer marketing published in Frontiers in Communication, credibility, trustworthiness, and attractiveness of an influencer emerge as three most investigated factors affecting marketing. Finally, financial reasons or the lack of them, and availability of the purchased products in stores are the two primary drivers of the purchase decision-making. Little is known about post-purchase communication patterns. Consumers tend to generate pre-purchase more frequently than post-purchase communications resulting in silent consumption (Bui et al., 2025). Zhang and Abdullah (2025) demonstrate the effect of digital content marketing on customer loyalty through emotional commitment. The term “silent consumption”, introduced later in this paper (Section 6.2) to describe the gap between strong pre-purchase influence and weak post-purchase content generation, is likewise an interpretive label coined to summarize this study's own stage-level asymmetry rather than a construct with an independent measurement basis in prior eWOM literature; its conceptual boundaries and relationship to established constructs such as eWOM engagement warrant further theoretical development.

2.5 Short-form video, interactivity, and purchase behavior

Using the CAB-based chain mediation analysis technique, Li and Sun (2025) in Frontiers in Communication have established that perceived interactivity in videos results in intention to purchase through a series of mediators as empathy and immersion occur due to interactivity. This literature is practically relevant to TikTok's position in the Bahraini market, where usage share and purchase-influence scores are compared directly in the Results section (Section 5.3). In particular, “shoppertainment”, the strategy through which purchasing is triggered using video entertainment and interactivity in a video stream, is what is used by TikTok as its main approach to stimulate purchase behavior (Bernstein and Guo, 2025). As indicated by Bernstein and Guo (2025) in their choice model, social commerce platforms stimulate impulsive buying decisions by optimizing product display and creating attention as well as social proof.

2.6 Credibility, trust, and authenticity in social media advertising

Credibility and authenticity have been proven to be important precursors to effectiveness in social media advertisements. For instance, according to Kothari et al. (2025), in their study they found out that credibility and sustainability were significant determinants of consumer behavior where trust acted as a partial mediator in the process. Such findings further support the importance of authentic communication between firms/brands and consumers (George et al., 2025; Sokolova and Kefi, 2020). Additionally, any relatively high influence attributed to alternatives' evaluation could plausibly be explained by the trust transferred from sources to recipients, which according to George et al. (2025) is one of the most common theoretical views used in social media influencer marketing studies; this stage-level pattern is examined empirically in Section 5.2. According to the systematic review carried out by Sardar et al. (2025), there are three important channels through which influencers' credibility affects consumer intentions to buy, namely (a) transfer of expertise in that the uncertainty level of consumers would decrease due to domain expertise, (b) trustworthiness in terms of building up a parasocial relationship, and (c) attractiveness whereby physical attractiveness acts as an important peripheral ELM cue affecting consumers irrespective of message strength. In this regard, in Bahrain, influencers are involved in promoting products through expertise, trustworthiness, and attractiveness, which offers a theoretical account for Instagram's platform-level influence as reported in Section 5.3.

2.7 Demographic moderators and the homogenization hypothesis

Classical consumer behavior theory predicts significant behavioral heterogeneity across demographic groups (Dukes and Liu, 2023). However, emerging scholarship suggests algorithmic personalization may be eroding these boundaries. Abdalla et al. (2025) found no significant age-based differences among Saudi social media users, and Kannan and Li (2017) theorized that digital marketing disrupts income-based segmentation through hyper-personalized delivery independent of socioeconomic status. While these studies address personalization effects in isolation, no prior GCC study has simultaneously tested all four major demographic moderators—gender, age, education, and income—within a single design. The present study provides the first such four-variable test in the GCC. Unlike Kannan and Li's (2017) personalization thesis, which concerns content-delivery mechanisms, the “homogenization hypothesis” advanced here is a behavioral proposition: that algorithmic curation renders demographic categories non-predictive of commercial susceptibility, with direct implications for audience segmentation strategy and consumer protection policy.

2.8 GCC and Bahraini consumer context

The GCC region constitutes an interesting case study for researching digital consumer behavior due to its distinct characteristics. With one of the highest global rates of social media penetration (full statistics in Section 1; DataReportal, 2024), platforms are present in the lives of all generations and income levels in Bahrain. In a market where Islamic consumption ethics, collectivist social norms, and income heterogeneity characteristic of hydrocarbon-based economies intersect with high-penetration digital infrastructure, purchasing behavior patterns may differ from those documented in Western contexts in theoretically meaningful ways. The absence of comprehensive research on social media's purchasing influence at different stages, especially in Bahrain where empirical evidence is lacking, poses the main research question of this work.

The literature review reveals several consistent ideas: (1) social media exerts stage-differentiated influence, with effects in the evaluation stage being significantly more impactful; (2) visual design and shoppertainment processes mediate marketing influence irrespective of demographics; and (3) the post-purchase electronic word-of-mouth communication behavior is least theorized. However, Western studies (Dukes and Liu, 2023) note a demographic impact on the influence of social media marketing, while studies conducted in the GCC (Abdalla et al., 2025) show moderation of the effect in advanced algorithmic markets—this conflict serves as the basis of this research.

Bahrain faces an emerging digital market infrastructure, which includes a mobile commerce approach, super-apps, and governmental initiatives on development of the digital economy, leading to an integrated social commerce environment. As shown by Foroudi et al. (2017), digital technology capability modifies the marketing management practices in emerging economies, especially concerning small businesses leveraging social media for competition. In Bahrain, the capabilities provided by Instagram such as product catalogs, shopping feature, and reel discovery fit well into the country's e-commerce plans according to the Bahrain Economic Vision 2030 (DataReportal, 2024). Such an analysis shows that stage-differentiated evidence is valuable because knowing the influence on consumer purchasing behavior isn't enough.

2.9 Social commerce architecture and the shortened purchase funnel

The integration of social networking, content publishing, and transactions in one platform interface that Hajli (2015) refers to the existence of the so-called “social commerce funnel”, an abbreviated form of the customer buying process during which all three steps take place without having the consumer leave the platform. According to Liang and Turban (2011), this abbreviated nature is exactly what distinguishes social commerce from traditional e-commerce: social activities (like sharing, reviewing, and recommending) and commercial activities (such as browsing, evaluating, and buying) complement each other. The significance of this abbreviation lies in its effect on the amount of cognitive work required by consumers. As stated previously, on Instagram, for example, consumers can see a product advertised via Reels, read users' reviews in the comments section, compare similar products in influencers' posts, and make a purchase using in-app checkout—in the same session. This structural feature leads to the theoretical expectation that Evaluation of Alternatives would register comparatively strong social media influence relative to the other stages, a prediction examined empirically in Section 5.2. If supported, this would suggest that social commerce platforms should be designed to prioritize evaluation of alternatives by providing rich comparison data in the context when it is needed most.

2.10 Impulse buying and algorithmic triggers in social commerce

Impulse shopping, where there is an absence of decision-making prior to the purchase made and whose action happens as a result of environmental cues, is becoming increasingly prevalent via social media thanks to the architectural attributes of these platforms that influence the finite attention span and emotional aspects of their consumers (Bernstein and Guo, 2025). Scarcity cues, social proof cues, and ambient entertainment are all used to make a trigger-based system for making purchases by incorporating them into commercial user interface design that differs from the traditional offline and online environments. Thus, Tiktok's shoppertainment approach and the product tags available in Instagram's Stories represent the architecture of impulse trigger delivery in Bahrainian social commerce. Yet, this reasoning suggests that Purchase Decision triggers may exert comparatively less influence than earlier deliberative stages, with a degree of residual rationality emerging once the actual purchase is made – a pattern examined empirically in Section 5.2 and interpreted, if supported, in light of Abdalla et al.'s (2025) findings on the mediating role of social media interactions in consumer decisions.

3 Research questions, objectives, and hypotheses

This section formally presents the study's primary and sub-level research questions (RQs) and their corresponding empirical hypotheses (H). Research questions are descriptive or exploratory in nature; hypotheses are testable directional statements subjected to formal inferential statistical testing. This distinction is maintained throughout.

3.1 Primary research question

RQ: What role do social media platforms play in shaping the purchasing behavior of consumers in the Kingdom of Bahrain?

3.2 Sub-Questions and hypotheses

RQ1 (S-O-R Stimulus → Response): To what extent does social media content exposure influence each stage of the consumer purchasing behavior process (Need Recognition, Information Search, Evaluation of Alternatives, Purchase Decision, Post-Purchase Behavior) among Bahraini consumers?

H1 (Tested via Pearson correlation; predicts a positive significant association): A statistically significant positive relationship exists between the level of social media content exposure and each of the five stages of purchasing behavior among consumers in the Kingdom of Bahrain.

RQ2a (Usage Frequency): Which social media platforms are most frequently used by Bahraini consumers in their purchasing-related activities?

RQ2b (Purchase Influence): Which social media platforms exert the greatest influence on Bahraini consumer purchase decisions?

H2 (Tested via One-Way ANOVA; predicts significant between-platform differences): Statistically significant differences exist in the degree of influence that different social media platforms exert on Bahraini consumer purchase decisions.

RQ3 (Organism-level moderators): Are there statistically significant differences in purchasing behavior attributable to demographic variables (gender, age, educational level, household income)?

H3 (T-test for gender; ANOVA for age, education, income; bidirectional test — classical theory predicts significant differences, homogenization hypothesis predicts null): H3 thus functions as a bidirectional test of competing theoretical predictions. Statistically significant differences exist in the level of social media's influence on purchasing behavior attributable to demographic variables.

4 Materials and methods

4.1 Research design

This study involved human participants: data were collected directly from 384 adult Bahraini social media users via a self-administered electronic survey. The study adopts a quantitative, cross-sectional, descriptive-analytical design. A survey methodology was employed for measuring attitudinal and behavioral variables across a large, geographically distributed sample of social media users. The study adheres to the ethical principles for human subjects' research, including informed consent, data anonymization, and voluntary participation. Ethical clearance was obtained from the University Research Council at Gulf University, Kingdom of Bahrain. All participants provided written informed consent prior to participation. No personally identifiable data were collected or retained. The study protocol complied with the Declaration of Helsinki (2013 revision) and applicable Bahraini data protection regulations.

4.2 Population and sampling

The target population was composed of all Bahraini residents aged 18 and above who make use of at least one social media platform for product/service discovery purposes as well as making informed purchase decisions. Non-probability convenience sampling was adopted, with an aim of achieving approximate stratification on several factors including gender, age range, level of education, and income group. The questionnaire was distributed electronically through social media platforms, university and professional networks, and community groups, using a snowball approach in which initial respondents were invited to share the survey link further; quota targets by gender and age band were monitored during collection but not strictly enforced, which is reflected in the demographic skew discussed below. A sample size of 384 respondents was obtained from Yamane's (1967) formula (confidence interval = 95%; margin of error = ± 5%) which is adequate for conducting multivariate statistical analyses (Hair et al., 2019). Post hoc power analysis using G*Power 3.1 software found sufficient statistical power for the one-way ANOVA test with k = 6 groups and N = 384 respondents with power = .99 for detecting a medium effect size (f = .25) with α = .05; and Pearson correlation analysis had power = .96 for detecting a small effect size (r = .15) with α = .05. Even though convenient sampling may not guarantee probabilistic generalizability, the current methodology is widely practiced in investigating consumer behavior using social media in GCC countries (Abdalla et al., 2025). Given the lack of a complete sampling frame for Bahraini social media users, convenience sampling in such a scenario was justified. However, there were three notable factors concerning sample attributes which limit the generalizability of the study findings. To begin with, the current sample favors female participants (60.9% vs. 39.1%). The proportion of female respondents is higher than the proportion of female Bahraini internet users (approximately 50/50). The second limitation concerns the over-representation of the young generation; 70.8% of the respondents were below the age of 35. This would lead to under-estimation of social media influence patterns for the mature population. Lastly, the lack of a complete sampling frame prevents any attempt to calculate the coverage error and design effect for the survey. Overall, the sample represents young and digitally savvy females in Bahrain who use social media.

4.3 Instrument

Data were collected via a structured electronic questionnaire in 2025, comprising: (1) demographic characteristics; (2) social media platform usage ranking; (3) a 9-item scale measuring social media influence across the five purchasing stages; and (4) supplementary platform-specific influence items. All attitudinal items used a five-point Likert scale (1 = Never; 5 = Always). Content validity was established through expert panel review by three specialists in marketing communication and digital media. Construct validity was confirmed via exploratory factor analysis on pilot data (n = 40), yielding a five-factor structure corresponding to the five purchasing stages. Internal consistency reliability: Cronbach's α = 0.87, exceeding the α ≥ 0.70 threshold (Nunnally, 1978). Discriminant validity was assessed by confirming that each factor correlated more strongly with its intended construct than with other factors; the average inter-factor correlation was .31, indicating acceptable construct differentiation (Hair et al., 2019). Stage-level means reported in Results were computed as unweighted averages of the item(s) assigned to each subscale (e.g., IS mean = mean of IS1 + IS2; EA mean = mean of EA1 + EA2). The Need Recognition subscale comprises a single item (NR1); as a result, subscale-level reliability (Cronbach's α) cannot be estimated for this dimension. For the remaining two-item subscales (Information Search, Evaluation of Alternatives, Purchase Decision, Post-Purchase Behavior), only the overall 9-item Cronbach's α (.87) is reported here; subscale-level reliability estimates were not computed for the present analysis and their absence limits confidence in the measurement structure at the individual-stage level, given the small pilot sample (n = 40) underlying the factor solution.

4.4 Data analysis

Data were analyzed using IBM SPSS Statistics (v. 27). Descriptive statistics (means, SDs, frequencies) characterized the sample. Inferential analyses included: (a) Pearson product-moment correlation coefficients to test H1; (b) One-Way ANOVA with Tukey HSD post hoc comparisons to test H2 (platform differences) and H3 components (age, education, income); (c) independent samples t-tests with Levene's test for equality of variances to test H3 (gender). Statistical significance threshold: α = 0.05. Effect sizes were computed for all inferential tests: Cohen's d for independent-samples t-tests; η² and ω² for ANOVA designs, following Cohen's (1988) benchmarks (small: η² = .01; medium: η² = .06; large: η² = .14). Statistical power was confirmed via post hoc analysis (G*Power 3.1) as reported in Section 4.2.

5 Results

5.1 Sample descriptive statistics

A total of 384 valid questionnaires were included in analysis (response rate = 91.4% from 420 distributed). Table 1 presents the full sociodemographic profile of the sample.

Table 1. Sample sociodemographic profile (N = 384).

5.2 RQ1/H1: social media influence across the five purchasing stages

Table 2 presents descriptive statistics for the five purchasing behavior stages, ranked by mean influence score. In Table 3, “social media exposure” refers to the overall 9-item influence-scale score described in Section 4.3 (i.e., the same composite measure underlying the stage-level means in Table 2), rather than a separate frequency-of-use measure.

Table 2. Descriptive statistics for five purchasing behavior stages (N = 384).

Table 3. Pearson correlation coefficients: social media exposure and five purchasing stages (N = 384).

r = Pearson coefficient; CI = 95% confidence interval via Fisher z-transform; all p < .001 (two-tailed); N = 384. Consistency r between stage means and r-values = .991. H1 fully supported.

The Evaluation of Alternatives stage recorded the highest mean (M = 3.54, SD = 0.98). Information Search ranked second (M = 3.43, SD = 1.06). Need Recognition scored M = 3.39, SD = 1.01. Purchase Decision influence was moderate (M = 3.05, SD = 1.04). Post-Purchase Behavior registered the lowest mean (M = 2.48, SD = 1.26), revealing the “silent consumption” asymmetry.

This combination of results constitutes the proof of Hypothesis 1. The positive correlations between the frequency of social media exposure and all five stages of consumer decision-making were established using the Pearson correlation analysis (Table 3). The most powerful correlations were found to be associated with the Evaluation of Alternatives (r = .487, p < .001), whereas the least significant correlation was obtained in relation to the Post-Purchase Behavior (r = .241, p < .001). It is consistent with the predictions made in accordance with the stimulus–organism–response model, where differential stimulus produces different engagement levels. Moreover, the monotone decrease of r from .487 to .241 coincides with the descending order of mean values (.487 > .453 > .421 > .374 > .241, with values 3.54–2.48 respectively), thus verifying the coherence of these data (r between means and correlations = .991). For Pearson correlations, effect sizes varied from low (r² = .058 for Post-Purchase Behavior) to moderate (r² = .237 for Evaluation of Alternatives). Thus, it can be concluded that, despite social media exposure having a significant effect on all stages, it is most predictive of deliberative, comparison-oriented behavior and least useful when predicting the content generated after purchase.

5.3 RQ2/H2: platform usage and purchase influence

Table 4 summarizes platform-level data comparing usage frequency rankings with mean influence scores.

Table 4. Social media platform usage frequency and purchase influence scores (N = 384).

Instagram demonstrated unambiguous dominance (53.3%; M = 4.21), consistent with its integrated social commerce architecture in the GCC region (Kothari et al., 2025). A noteworthy divergence exists between TikTok's usage rank (third, 17.5%) and its influence score (M = 3.85, second highest), consistent with Li and Sun's (2025) CAB model and the shoppertainment architecture described by Bernstein and Guo (2025). Facebook and X recorded minimal usage and influence (M = 2.12 and 2.45, respectively). One-Way ANOVA (Table 5) confirmed statistically significant differences in purchase influence scores across the six platforms, F(5, 378) = 12.91, p < .001, η² = .146, indicating a large effect (Cohen, 1988). Tukey HSD post hoc pairwise comparisons (Table 6) revealed two distinct clusters: high-visual platforms (Instagram, TikTok, Snapchat) each significantly outperformed text-dominant platforms X and Facebook (all p < .05). H2 is fully supported. These platform-level comparisons should nonetheless be interpreted with caution given the highly unequal group sizes underlying them: the Facebook (n = 4) and X/Twitter (n = 6) cells are far smaller than Instagram (n = 205), which raises concerns about the stability of these group means and the reliability of pairwise comparisons involving the two smallest categories, notwithstanding Tukey HSD's use of pair-specific harmonic-mean sample sizes (Kramer, 1956) to partially compensate for unequal n. Levene's test results for homogeneity of variance across the six platform groups are not separately reported for this comparison and should be checked before the ANOVA assumptions are considered fully satisfied; the comparisons involving Facebook and X in particular should be treated as provisional pending replication with larger, more balanced platform-level samples.

Table 5. One-way ANOVA: between-platform differences in purchase influence scores (N = 384).

F(5,378) = 12.91, p < .001; η² = .146 (large effect, Cohen 1988); ω² = .134. Two-cluster structure: visual platforms (Instagram, TikTok, Snapchat) vs. text-dominant (YouTube, X, Facebook).

Table 6. Tukey HSD post hoc pairwise comparisons: platform purchase influence (N = 384).

ns, not significant. HSD computed with pair-specific harmonic mean of n (Kramer, 1956). q-critical (k = 6, df = 378, α = .05) = 4.03.

p < .05.

p < .001.

5.4 RQ3/H3: demographic differences in purchasing behavior

Table 7 shows the inferential test results with effect sizes: demographic differences in purchasing behavior of the study sample.

Table 7. Inferential test results with effect sizes: demographic differences in purchasing behavior (N = 384).

Effect sizes added (Cohen, 1988): d for t-test (d = |M₁−M₂|/SD_pooled); η² for ANOVA (SS_Between/SS_Total). All effects negligible–small, confirming demographic convergence.

5.4.1 Gender (H3a)

Independent samples t-test revealed no statistically significant difference between male (M = 3.21, SD = 0.95) and female (M = 3.23, SD = 1.02) respondents, t(382) = 0.046, p = 0.963, Cohen's d = .02 (negligible effect). Levene's test for equality of variances was non-significant, F = 1.84, p = .175, confirming the homogeneity of variance assumption was met. H3a is rejected (null retained). See Table 8.

Table 8. Independent samples T-test results by gender (N = 384).

Cohen's d = .02 (negligible). Levene's F = 1.84, p = .175 (homogeneity of variance assumption met).

5.4.2 Age (H3b)

One-way ANOVA comparing purchasing behavior across five age cohorts yielded F(4, 379) = 2.015, p = 0.093, η² = .021 (small effect), a non-significant result. H3b is rejected (null retained). See Table 9.

Table 9. One-Way ANOVA for Age differences in purchasing behavior (N = 384).

η² = .021 (small effect). Post hoc not applicable (non-significant F).

5.4.3 Education and household income (H3c, H3d)

Parallel ANOVAs for educational level, F(3, 380) = 0.463, p = 0.708, η² = .004 (negligible), and household income, F(3, 380) = 1.150, p = 0.328, η² = .009 (negligible), yielded non-significant results. Both H3c and H3d are rejected (null retained). Table 10 presents the detailed ANOVA outputs for education and income sub-hypotheses.

Table 10. One-way ANOVA results: education level and household income (N = 384).

η² (Education) = .004 (negligible); η² (Income) = .009 (negligible). Post hoc tests not applicable (non-significant F-values). Effect size benchmarks per Cohen (1988).

Null result for education requires further theoretical considerations and discussion. According to advertising literacy theory proposed by Boerman et al. (2017), a person who received more education has better skills to defend against the persuasive impact of advertisement due to improved awareness and cognitive counter-arguments to manipulation attempts. Nevertheless, results of the current research show that this defense system works poorly in a social media environment where emotional visual communication prevails. This result can be predicted using Elaboration Likelihood Model, which states that educational attainment contributes to the central route of persuasion and does not have significant importance when affective processing takes place instead of analytical processing of stimulus. Fransen et al. (2015) provide additional support to this argument by demonstrating that people will start resisting manipulations only in case when they are mentally able and motivated to analyze the arguments of the persuasive message and activate counter-arguments. In other words, when an environment with many peripheral cues predominates, people cannot rely on their skills to analyze the message to become less vulnerable to manipulation attempts. Therefore, platform regulation can be recommended because conventional consumer protection through media literacy programs appears insufficient under conditions where persuasion operates via peripheral channel. As the example, the disclosure of commercial stimuli and transparency of algorithms should be considered (Evans et al., 2017).

Null result concerning income can be interpreted from a theoretical perspective as well. Income is known to serve as a major factor in traditional economic models, which postulate that people with different incomes will act differently while consuming goods and services. Non-significant ANOVA test result [F(3, 380) = 1.150, p = .328, η² = .009] implies that social media susceptibility to commercial stimuli is evenly distributed among various levels of income. This study uses the term “consumer democracy” as a working label – not an established construct from prior theory – to describe an income-independent pattern of commercial susceptibility; it is offered as an interpretive summary of the present null result rather than a validated theoretical concept. This finding seems justified considering the structure of social media content delivery that is based on personalized recommendations irrespective of people's incomes. Moreover, platforms' costs remain so low that they can be accessed by most people regardless of their financial status (DataReportal, 2024). Kannan and Li (2017) state that in terms of personalized advertising, the impact of digital marketing on traditional income segregation can hardly be overestimated since income becomes irrelevant for effective content recommendation. Therefore, results related to income in the Bahraini sample do not support Hypothesis 3 (H3).

6 Discussion

6.1 Cognitive-evaluative primacy and the hierarchy of stage effects

The existence of hierarchy of influence in social media platforms, more evident during Evaluation of Alternatives (r = .487) and Information Search (r = .453), than Purchase Decision (r = .374), is arguably the most important result of this study from a theoretical perspective. This phenomenon differs from the transactional paradigm regarding social media influence, as it empirically supports the findings of Wan et al. (2025) about how social media participation mainly impacts on purchase intentions in terms of self-disclosure and social interaction, as opposed to commercial influence. The rationale provided by the S-O-R approach for explaining this result is quite simple: organism-based processes, which include comparative perception, risk management, and credibility evaluation, become most activated at evaluation time, thereby making the electronic word-of-mouth stimulus the most effective. From an applied standpoint, this result has obvious managerial implications for practitioners who decide to invest their social media efforts in comparison-based content formats, since they can expect greater business outcomes than direct purchase-related materials.

6.2 Silent consumption and the eWOM structural Gap

The marked disparity between pre-purchase engagement levels (r = .421–.487) and post-purchase contribution (r = .241) constitutes empirical evidence of what this study terms “silent consumption”: consumers are actively influenced by social media during deliberative stages but rarely reciprocate by generating post-purchase content. Possible explanations include privacy concerns associated with Bahraini consumer culture, the lack of review incentives at the platform level, and satisfaction imbalance (Sokolova and Kefi, 2020). Indeed, Zhang and Abdullah (2025) found out that digital marketing content affects customer loyalty because of emotional commitment mechanisms, which would drive customers to create post-purchase content if they existed. Thus, the aforementioned information asymmetry creates a structural problem for Bahraini social commerce ecosystem. Its impact can be observed in the form of the commercial impact of the information gap, as when social media have a decisive effect on the evaluation process (Cheung and Thadani, 2012), but people do not produce any review posts in return. As a consequence, there would be a situation wherein more users mean less evaluative content produced due to the lack of contribution, creating a self-limiting information gap problem. To counteract this problem, brands could use post-purchase gamification strategies, community-based recognition, and post-purchase activation campaigns that stimulate emotional commitment mechanisms, as noted by Zhang and Abdullah (2025).

6.3 Platform differentiation: Instagram's dominance and TikTok's disproportionate influence

In the case of Instagram (53.3%; M = 4.21), the near monopoly represents how well the commercial architectural aspect of marketing blends into the Bahraini consumers' sense of aesthetics. As observed in the marked difference in the effect and actual use of TikTok, the theory provided by Li and Sun (2025) on the Commercial Architecture of Behavior (CAB) clearly explains that the use of short video clips may produce purchase intention among users regardless of the presence of prior product interest based on the S-O-R stimulation theory. Based on the ELM's peripheral route to persuasion, the success of TikTok in terms of marketing comes from emotions evoked during consumption of content rather than any logical reasoning involved in understanding its content.

6.4 Demographic convergence as evidence of algorithmic dominance

The non-significant results from all four demographic moderators – gender, age, education, and income – counter an assumption inherent to demographic-based audience segmentation. According to classical advertising theory, such demographic factors should generate significantly different response patterns (Dukes and Liu, 2023). However, our results fail to confirm this assumption for the Bahraini social media environment. In conjunction with the Saudi findings reported by Abdalla et al. (2025) regarding non-significant age-based moderation, both sets of results support a regional trend that we refer to as “demographic convergence”. One plausible explanation for this convergence is algorithmic personalization: recommendation engines may provide commercial messages to users somewhat independently of demographic profile, which could equalize exposure that would otherwise vary by demographic group. This explanation is consistent with the pattern of null results but was not directly tested, since the study did not measure algorithmic delivery processes; it is offered here as an interpretation rather than a demonstrated mechanism. The algorithmic convergence hypothesis does not necessarily rule out the statistical power hypothesis. Post hoc power analysis (Section 4.2) shows power ≈ .99 for medium effects, but ≈ .35 for very small effects (f < .10). Hence, very small differences between the groups might have been missed in the experiment. However, given that null results were obtained across four theoretically independent demographic moderators each associated with its own classical prediction, it seems more plausible to attribute them to the convergence phenomenon rather than insufficient power on all fronts. Two other hypotheses, however, merit mention here. One possibility is sample demographics: the present sample is 60.9% female and 70.8% under the age of 35, thus making it less diverse on both counts, especially on the latter one (there are only n = 9 people aged 55 or above). There is good reason to suspect that the failure of demographic factors to have any effect on the results is at least partly attributable to limited variation between groups as opposed to, or in addition to, true homogenization of algorithms. The second possibility concerns insufficient sensitivity at the composite scale level: the total scale average was used as the dependent measure in the demographic ANOVA analyses. Differences between individual stages may exist even if the overall composite score stays the same; thus, an additional ANOVA test of each stage in turn against four demographic variables is advisable.

6.5 Practical implications for marketers, regulators, and scholars

This study has implications for three different stakeholders. First, marketing practitioners are advised to reassign budget allocation from triggering content for purchase to deliberative content such as product comparisons, aggregate user reviews, and influencer demonstrations that communicate instrumental value. The finding of the interaction between informational value and entertaining value established by Kim and Ko (2012) shows that a content strategy focused on deliberation rather than promotion can boost customer equity gains. Second, the demonstration by Hollebeek et al. (2014) of cognitive, affective, and behavioral dimensions of brand engagement shows that evaluation-stage campaigns that induce all three outcomes are associated with more stable purchase intention formation than campaigns limited to affective reactions.

Third, social media commercial influence poses a consumer protection problem that current regulatory frameworks cannot resolve effectively because of demographic convergence. Policies that focus only on protecting the youth, low-income, and less-educated consumers leave most consumers vulnerable when social media commercial influence cuts across demographics equally. Voorveld et al. (2018) found that type of platform rather than demographic characteristics determines engagement in social media advertising. While disclosure increases advertisement recognition on Instagram, as shown by Evans et al. (2017), we find that there was no effect of education level on recall, meaning that educated individuals also can be influenced via peripheral processing routes. Thus, our results support the conclusion that platform-level policies should govern social media commercial influence, as Shareef et al. (2019) found platform architecture as a moderator of advertising effectiveness. Policy tools based on mandatory labeling of commercial content, transparency requirements, and platform-level disclosure should be used to combat social media commercial influence.

For researchers in the field, the multi-model integrative approach of three layers (UGT + S-O-R + ELM) shows that no single theory explains social media commercial influence. UGT is responsible for explaining why certain platforms are chosen by consumers, whereas S-O-R is responsible for explaining how stimulus causes a stage-level response in consumers, and ELM explains whether instrumental processing capacity moderates susceptibility to commercial content. Using all three together yields more accurate and falsifiable predictions compared to models relying on one theory. Zhu and Zhang (2010) demonstrated that consumer characteristics can interact with online review influences, but their study used a framework that did not explain this effect; our integrative approach addresses this concern. The impact of influencer credibility on consumer purchase intentions through belief systems (organism-level phenomenon) was identified by Lim et al. (2017).

7 Conclusions

Among the surveyed sample – predominantly younger and female relative to Bahrain's general population (see Section 4.2) – findings show that social media platforms have embedded themselves as a prominent part of the consumer journey, shaping purchase-related behavior in differentiated ways across the five decision-making phases. These conclusions should be read as characterizing this sample rather than the Bahraini consumer population as a whole. Social media platforms are shown to be primarily used as a cognitive evaluation resource among Bahraini consumers; these platforms help users in reducing the uncertainties related to their purchasing behavior, demonstrated by high Pearson correlations for Evaluation of Alternatives (r = .487) and Information Search (r = .453). Using the three layers of UGT + S-O-R + ELM theoretical framework, Instagram was found to be the prominent commercial platform driven by aesthetic gratification (UGT), visual stimulus response process (S-O-R) and peripheral route persuasion (ELM). On the other hand, the strong persuasion capabilities of TikTok were identified, driven by shoppertainment through peripheral route processing. Absence of any significant demographic moderating effects in terms of gender, age, education and income levels – all with negligible to small effect sizes – offer empirical proof of demographic convergence.

Implications can be drawn for various stakeholders on account of this study. Marketing professionals must take cognizance of the evaluative primacy result, which recommends investment in content in social media platforms that allow comparisons such as side-by-side product evaluations, amplified consumer reviews and influencer demonstrations instead of direct purchases. In view of the finding of the silent consumption, there should be an emphasis on the design of incentive architecture that encourages post-purchase review processes such as review gamefication and reward systems to address the eWOM gap and its consequences. Platform regulators will find the evidence of need generation, shoppertainment-based impulses and demographic convergence of consumer vulnerability as grounds that necessitate considerations of disclosure requirements and consumer protection measures not covered by existing GCC frameworks. Finally, communication researchers will benefit from the three-layer UGT + S-O-R + ELM framework, stage-differentiated correlation analysis, and demographic convergence as advances made towards a mechanism-based understanding of digital consumer behavior.

Among the limitations identified are the convenience sampling technique employed that makes generalizations about the population statistically tenuous, the cross-sectional survey design that rules out cause-effect inference, self-report bias and weak statistical power to identify very small demographic effects. There are four directions for future research: (1) experimental designs involving random assignment of Bahraini consumers to condition (Instagram, TikTok or a text-based control group) for establishing the causal effect of the differential stimulus-response (S-O-R) prediction; (2) equivalence testing designs utilizing TOST procedure with pre-registered minimum threshold effect sizes (d = .20) for establishing whether the four null demographic results reflect the truth versus Type II error; (3) cross-national GCC study of Qatar, Oman and Saudi Arabia comparing demographic convergence of consumer vulnerability in markets with lower social media adoption rate, following the same methodological approach and enabling coefficient comparison; and (4) longitudinal design incorporating digital trace methodology combining self-report measures with actual platform transaction records, and thus providing causal inference regarding the chain of events from social media exposure to purchase decision and testing whether TikTok's influence-to-usage ratio (M = 3.85 at 17.5% penetration) would hold under controlled exposure conditions.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fcomm.2026.1897772/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Steven Bellman, Adelaide University, Australia

Reviewed by: Claudiu Coman, Transilvania University of Brașov, Romania

Lana Sularto, Gunadarma University, Indonesia