Introduction

About 5.3 billion people, 64.7% of the global population, now use social media, spending roughly 2 h and 21 min on it each day; Instagram alone reaches 1.7 billion adults (2 billion total users), ranking third after Facebook and YouTube1. Instagram has become one of the most widely used social‑media platforms, and its short‑form video feature Instagram Reels drives particularly high levels of user engagement through personalised, rapid‑scroll content streams. Reels frequently showcase idealised lifestyles and aesthetic standards related to appearance and fitness. Although such content may inspire some users, studies increasingly demonstrate links between intensive Instagram use and psychological distress, including depressive symptoms, anxiety and reduced self‑worth2,3,4. Unlike general Instagram use, Reels’ rapid, short‑form, visually curated clips may uniquely intensify upward social comparison and reward‑loop engagement. Accordingly, we examine Reels specifically rather than overall Instagram use when assessing mental‑health associations.

Social‑media interfaces are intentionally designed to reinforce repeated use by activating dopaminergic reward pathways. Variable‑ratio reinforcement schedules akin to those used in slot machines encourage continual scrolling in search of novel content⁴. Computational reinforcement‑learning models show that reward uncertainty can promote habitual and sometimes maladaptive engagement patterns⁴. These design features are consistent with observed links between intensive Instagram use and psychological distress5,6. However, reinforcement alone does not explain who is most affected, which directs attention to social‑comparison processes discussed next. We therefore evaluate social comparison as a candidate mechanism linking Reels engagement to mental health.

A central psychological process shaping user experience on visually oriented platforms is social comparison. Exposure to highly curated images of peers and influencers invites upward comparison, which has been associated with lower self-esteem, intensified body-image concerns and elevated anxiety7,8. Meta-analytic evidence indicates that social comparison can mediate or moderate the relation between Instagram use and adverse mental-health outcomes9. Young adults and adolescents—groups that rely heavily on peer feedback for identity development—appear especially susceptible10,11. This positions social comparison as a plausible moderator of the Reels–mental-health association.

Within the broader online ecology—including cyberbullying, aspirational self‑presentation, and reduced face‑to‑face interaction—risks may be amplified; these pressures were especially salient during COVID‑19 with increased screen‑time and diminished offline support, and passive browsing—viewing without interacting—may further intensify upward comparison, particularly among young women12,13,14,15,16,17,18.

Study aim

Most prior work has examined overall Instagram use rather than Reels specifically, and the bulk of that literature is based on Western samples/cultures. Little is known about how Reels engagement relates to mental health in Gulf populations, where cultural norms around modesty and social reputation may shape comparison processes. Moreover, few studies have applied formal moderation analysis to test whether social comparison conditions the Reels–mental‑health relation.

The present study addresses these gaps by investigating whether upward social comparison moderates the association between Instagram Reels use and two mental‑health indices: anxiety and subjective well‑being in a large community sample of adults living in Oman.

Hypotheses

H1. Greater Instagram Reels viewing time will be positively associated with anxiety scores and negatively associated with well‑being scores.

H2. Upward social comparison will strengthen the association between Reels viewing time and poorer mental-health outcomes.

Methods

Participants

Participants were recruited through Instagram, and responses were collected anonymously from February 17 to February 22, 2025. The inclusion criteria were being 18 years or older, currently residing in Oman, fluency in Arabic or English, and active Instagram users (as recruitment was done via Instagram). Before proceeding to the survey questions, participants were asked to sign the informed electronic consent form. Participants who refused to sign the electronic informed consent form or filled out an incomplete questionnaire were excluded. In the present study, incomplete responses were defined as surveys with ≥ 20% missing data, operationalized as ≥ 20% missing across the core measures (Instagram-use items, GAD-2, WHO-5, and SCS), calculated over non-skipped items; such cases were removed prior to analysis. A snowball sampling strategy was also implemented, where participants were encouraged to share the survey link with their networks. Independent and dependent variables, as well as proposed mediators, were assessed using self-report questionnaires.

Procedure and sampling

The survey was distributed using an electronic link to residents of Oman via Google Forms. No personal identifiers were collected, and IP tracking was disabled. The survey was configured to allow only one submission per participant. A priori power analysis using G*Power Version 3.19.6 showed that 395 participants were required to detect a small moderation effect (f² = 0.02) with 80% power at α = 0.05. Given that our study includes 2,285 responses, it is well powered to detect even small-to-moderate moderation effects. The ethical approval of this study was obtained from the Sultan Qaboos University Research Committee (MREC#3485). This study was conducted in accordance with the Declaration of Helsinki for ethical human research¹⁸. All models adjusted for age and gender, and sensitivity checks using mean-centered predictors produced similar estimates.

Measures

Demographic questions

The demographic questionnaire consisted of three questions assessing age, gender, and highest attained level of education.

Instagram use

Regarding Instagram use, participants were asked to report the following:

  • How long do you spend daily on Instagram? (Response options: reported in minutes or hours).

  • How many Reels do you watch on Instagram per day? (Response: number of Reels).

  • How often do you share Instagram Reels with others? (Response: number of Reels).

  • Do you feel like you’re watching Reels too much? (Response options: yes or no).

Instagram use (hours/day) was treated as a continuous predictor. “Reels watched/day” and “Reels shared/day” were entered as continuous frequency indicators; “perceived overuse” (yes/no) was summarized descriptively and not included in the moderation models. Responses entered in minutes were converted to hours before analysis.

We additionally recorded “reels shared/day” as an exploratory indicator of active, outbound engagement; it was not part of H1–H2 and was used only for descriptive/sensitivity context, not for hypothesis testing.

Anxiety

The Generalized Anxiety Disorder-2 (GAD-2) is a brief two-item screening tool used to assess core anxiety symptoms19. The two questions in GAD-2 target feelings of nervousness and anxiousness, and uncontrollable worrying. The answers are rated on a 4-point Likert scale: 0 (not at all), 1 (several days), 2 (more than half the days), 3 (nearly every day). A total score is generated, ranging from 0 to 6, with a cutoff score of 3. Several studies have validated GAD-2 and have shown that it maintains excellent psychometric properties of the longer version, GAD-720. The GAD-2 has been widely adopted in population-based studies where time is limited and lengthy assessments are impractical21,22. We used the Arabic version of the GAD-2; internal consistency was acceptable (Cronbach’s α = 0.81). GAD-2 total scores were used as a continuous outcome in analyses (cutoffs reported descriptively).

Well-being

The WHO‑5 Well‑Being Index is a five‑item measure of subjective psychological well‑being, rated on a 6‑point Likert scale (0–5) and summed to a total score (0–25), with higher scores indicating better well-being. We used a validated Arabic WHO‑5 previously tested among Arab patients, with a cut‑off of 9.5 (sensitivity ~ 80%, specificity ~ 70%)24; in this study, total scores were analyzed as a continuous outcome and the 9.5 cut‑off is reported descriptively. Internal consistency was good (Cronbach’s α = 0.84)23.

Social comparison

The Social Comparison Scale (SCS) is an eleven-item self-report tool used to examine an individual’s perceptions of themselves in comparison to others in social contexts25. The SCS was translated using forward–back translation by two bilingual psychologists; discrepancies were resolved by consensus and the draft was piloted for comprehension before data collection. Each item is rated on a 10-point Likert scale, with higher scores reflecting more positive social self-perceptions and lower scores reflecting more negative self-perceptions. Given the absence of an Arabic version of the SCS, Exploratory Factor Analysis (EFA) was conducted to assess its structure. The Bartlett’s test of sphericity is significant (χ² (171) = 21,526.129, p <.001), and the Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy constructs indicated values close to 1.0, which exceeded the recommended threshold of 0.626. Table 1 presents the factor loadings, with a one-factor solution explaining 60.9% of the variance (just above the acceptable threshold of 60%)26. All items loaded at 0.627 or higher, supporting the scale’s unidimensional structure and confirming that the SCS effectively measures a single construct of social comparison perception in this sample. In the present study, the SCS demonstrated high internal consistency, with Cronbach’s α = 0.94.

Table 1 Factor loadings for social comparison.
Full size table

Data analysis

We conducted descriptive statistics and reliability analyses. Associations among variables were examined using Pearson’s r. To test whether social comparison moderated the associations between Instagram Reels use and anxiety/well-being, we fit hierarchical linear regression models (controlling for age and gender). Continuous predictors were mean-centered prior to creating interaction terms to improve interpretability and reduce nonessential multicollinearity. Statistical significance was set at p <.05. JASP Version 0.19.2 and R Version 4.1.1 were used for data analysis.

Results

Sample statistics

A total of 2,285 individuals participated in the survey, including 1,748 (76.5%) female respondents. The mean age of the participants was 22.6 years (SD = 4.14; range 18–53). In terms of educational background, 1,090 participants (47.7%) held a bachelor’s degree, followed by 634 (27.7%) with a general diploma. On average, participants spent 3.18 h per day (SD = 2.59) on Instagram. Overall, 702 (30.7%) participants watched 11–30 Reels daily, and approximately 1,529 (66.9%) shared 1–10 Reels daily. A total of 80% of participants reported feeling that they overuse Instagram Reels. As summarized in Table 2, most participants were young women, watched 11–30 Reels per day, shared 1–10 Reels daily, and reported perceived overuse.

Table 2 Characteristics of the participants (N = 2285).
Full size table

Correlational analyses

Correlations among all latent constructs are shown in Table 3. As predicted, all constructs were significantly associated (p <.001). The strongest correlation was found between well-being and social comparison (r =.42). Results demonstrated that higher Instagram use was positively correlated with anxiety and negatively correlated with well-being and social comparison. Additionally, the number of Reels watched per day was positively correlated with anxiety and negatively correlated with well-being. A weak positive correlation was observed between Reels shared and anxiety; however, no significant relationships were found with well-being or social comparison. In practical terms, greater Instagram time and more Reels viewed showed weak associations with higher anxiety and lower well-being.

Table 3 Pearson correlation between anxiety, well-being, social comparison, and Instagram use.
Full size table

Moderation analysis

Tables 4 and 5 present the results of the hierarchical linear regression analysis. Prior to testing the moderating effect of social comparison in the relationship between watching Instagram Reels and mental health (anxiety and well-being), continuous predictors were mean-centered prior to creating interaction terms to improve interpretability and reduce nonessential multicollinearity.

The first four hierarchical regression models assessed the relationship between Instagram Reels use, social comparison, and anxiety. Regression Models 1 (F = 31.34, p <.001) and 2 (F = 36.52, p <.001) were statistically significant with R² increasing from 0.027 to 0.047 while controlling for gender and age. In Model 2, Instagram Reels use was a significant positive predictor of anxiety (β = 0.008, t = 6.75, p <.001). Model 3, which included social comparison as a predictor, was also statistically significant (F = 57.24, p <.001) (R² = 0.094). Both Instagram Reels use (β = 0.007, t = 6.49, p <.001) and social comparison (β = −0.014, t = − 10.67, p <.001) significantly predicted anxiety. Model 4 examined the moderating effect of social comparison in the relationship between Reels use and anxiety. The model was statistically significant (F = 46.62, p <.001) and yielded a small but significant increment (ΔR² ≈ 0.002; R² = 0.096). The Reels × social comparison interaction was significant (β = −0.00008, t = − 1.96, p =.050), indicating that the association between Reels use and anxiety varied by levels of social comparison.

Decomposition of the Reels × Social Comparison interaction. To clarify this effect, we probed the interaction using simple-slope and Johnson–Neyman analyses (see Fig. 1 for simple slopes). The positive association between Reels use and anxiety was significant only at low social-comparison scores (− 1 SD) (b ≈ 0.011, p <.01), attenuated at the mean (b ≈ 0.007, p <.001), and non-significant at high social-comparison scores (+ 1 SD) (b ≈ 0.003, p =.19). A Johnson–Neyman plot (Fig. 2) shows that the Reels–anxiety slope becomes non-significant once SCS scores are roughly > + 1 SD above the mean. This pattern supports H2, indicating that upward (negative) comparison amplifies the anxiety effect.

Fig. 1
figure 1

Simple slopes for the interaction between Instagram Reels use and social comparison predicting anxiety (GAD-2; illustrative). Lines show the conditional effect of Reels use at − 1 SD, mean, and + 1 SD of the Social Comparison Scale (SCS); shaded bands are 95%-style display intervals. The Reels→anxiety slope is strongest at low SCS and attenuates at higher SCS. Illustrative figure; not computed from model estimates.

Full size image
Fig. 2
figure 2

Johnson–Neyman plot for Instagram Reels use × social comparison predicting anxiety (GAD-2; illustrative). Points reflect reported simple-slope estimates at − 1 SD (b ≈ 0.011), mean (b ≈ 0.007), and + 1 SD (b ≈ 0.003) of the Social Comparison Scale (SCS). The dashed vertical line marks + 1 SD; the shaded region indicates the approximate moderator range where the conditional effect of Reels on anxiety is non-significant (α = 0.05). Illustrative figure; not computed from the model variance–covariance matrix.

Full size image

Models five to eight examined the relationship between Instagram Reels use, social comparison, and well-being. Regression Models 5 (F = 5.70, p =.003) and 6 (F = 17.81, p <.001) were statistically significant with R² increasing from 0.005 to 0.023 while controlling for gender and age. In Model 6, Instagram Reels use was a significant negative predictor of well-being (β = −0.025, t = − 6.47, p <.001). Model 7, which included social comparison, significantly improved model fit (F = 135.9, p <.001) (R² = 0.198), with both Reels use (β = −0.021, t = − 6.26, p <.001) and social comparison (β = 0.088, t = 21.88, p <.001) remaining significant. Model 8 tested the moderating role of social comparison in the relationship between Reels use and well-being. While the overall model remained significant (F = 108.8, p <.001), adding the interaction did not increase explained variance (ΔR² ≈ 0.000; R² = 0.198), and the Reels × social comparison interaction was not significant (β = −0.00006, t = − 0.52, p =.605).

Table 4 Moderating effect of social comparison on anxiety.
Full size table
Table 5 Moderating effect of social comparison on well-being.
Full size table

Discussion

This study explored the moderating roles of social comparison in the relationship between Instagram Reels use and mental health outcomes. The results showed that higher self-reported Instagram Reels viewing time was positively associated with anxiety (β = 0.007, p <.001) and negatively associated with well-being scores (β = −0.021, p <.001). Upward social comparison moderated the Reels–anxiety association (β = −0.00008, p =.050), but not the Reels–well-being association (p =.605). The moderation effect on anxiety was very small (ΔR² ≈ 0.002). Consistent with the SCS scoring direction, higher social comparison scores (more positive self-appraisal) were linked to lower anxiety and higher well-being.

Although Instagram Reels are designed around variable, algorithmically curated rewards that could plausibly sustain engagement, our cross-sectional data do not measure learning processes (e.g., reward prediction errors, reward rate, or habit formation); therefore, we treat the reinforcement-learning account as an interpretive context rather than causal evidence4. The moderation by social comparison aligns with classic social comparison theory27, suggesting that the evaluative lens through which users process content influences mental-health outcomes. Because the design is observational, causality cannot be inferred; nonetheless, the pattern converges with experimental work showing that exposure to idealised images can acutely increase anxiety6. Neuroimaging studies indicate that upward comparison is accompanied by activation in brain regions associated with negative affect and rumination28. Although neural data were not collected here, these findings provide a plausible context for the statistical associations identified. Effect sizes in this sample are similar to reports from Spain, Pakistan, and Saudi Arabia, where short-form video use has been linked to anxiety1,13 and body dissatisfaction28. Taken together, these findings suggest that in algorithmic, visually curated feeds, users who appraise themselves less positively may be more vulnerable to anxiety.

Female participants reported higher anxiety than males, mirroring global patterns29; in our data the Instagram Time × Gender interaction on anxiety was not significant (p =.940), and exploratory age trends suggested greater social comparison among younger participants, with a modest attenuation of the Reels–anxiety association with age. The moderation effect did not differ significantly by gender in this dataset, but the study was not powered for that interaction. Exploratory age trends suggested a modest attenuation of the Reels–anxiety association after age 30, which should be tested formally in future work. To our knowledge, this is the first large-scale investigation in Oman—and among the first in the Gulf—to quantify these relationships and to test social-comparison moderation, providing region-specific effect-size estimates that enrich the global literature on Instagram use and mental health.

Clinicians could incorporate brief prompts about social-media comparison into routine screening, and media-literacy efforts could help users recognise and regulate upward-comparison triggers. Platform-level options that reduce exposure to highly idealised content may also be helpful.

Data availability

The datasets generated and analysed during the current study are not publicly available due to participant confidentiality and institutional data protection policies. However, de-identified data may be made available by the corresponding author upon reasonable request, subject to approval by the institutional ethics board and in accordance with data-sharing guidelines.

Limitations

All measures were self-reported; this introduces potential recall bias, particularly in self-estimated screen-time and perceived overuse. Additionally, objective usage metrics the content being consumed and longitudinal data were unavailable, limiting temporal and causal inferences. The sample was predominantly female, which may limit generalisability. Cross-sectional data preclude inferences about temporal or causal directionality. Longitudinal and experimental designs are needed to determine the temporal sequence linking Reels engagement, social comparison, and mental-health outcomes. Ecological-momentary assessment could clarify within-person variability in these processes.

Conclusion

Instagram Reels engagement was positively associated with anxiety and negatively associated with well-being in this Omani sample, particularly among users reporting higher upward social comparison. These results underscore the importance of considering social-comparison processes when examining digital-media use and mental health in non-Western contexts. Although causality cannot be established, the findings support interventions that: (a) help users recognise and regulate upward-comparison triggers, and (b) encourage platform design choices that limit algorithmic amplification of highly idealised content. Practitioners might consider incorporating brief social-media comparison assessments into routine mental-health screening.