Associations between social media-based emotional support and internalizing symptoms: a longitudinal study in a national sample of U.S. adolescents and emerging adults

1 Introduction

The young people mental health crisis shows no signs of receding, and compared to their peers, adolescents and emerging adults (10–25 years of age) from minority backgrounds (i.e., race, ethnicity, LGBTQ+, or rural living) face higher risk for internalizing mental health symptoms, such as depression, anxiety, and loneliness (King et al., 2008; Hafeez et al., 2017; McDanal et al., 2021). At the same time, social media (SM) use among young people is ubiquitous in the U.S., where 75% of teens and emerging adults use at least one SM platform daily (Faverio and Anderson, 2025). Given that adolescence and emerging adulthood are sensitive developmental stages, the ubiquity of SM use raised concerns that it might negatively impact young people's mental health.

Over the last decade, research examining associations between SM use and young people's mental health has largely focused on time spent on social media. Findings have been inconsistent, with studies reporting small positive, null, or bidirectional associations of time spent on social media with outcomes such as depression, anxiety, and loneliness (Twenge et al., 2018; Heffer et al., 2019; Vuorre et al., 2021; Orben et al., 2022). Meta-analytic evidence suggests that these effects are, on average, very small and account for minimal variance in mental health outcomes, too small to warrant policy changes (Orben and Przybylski, 2019; Liu et al., 2022).

The small size of SM use effects on mental health suggests that time spent is a weak predictor of mental health outcomes, and the heterogeneity of these effects suggests that the field needs to keep investigating what specific exposures might be harmful or protective (and for whom), other than simply amount of use. It has also been suggested that interacting in rapidly changing SM platforms that are vastly different from in-person interactions (e.g., permanence of SM posts, asynchronicity of SM interactions, quantifiability of SM metrics) is the real driving force behind the effect of social media on health (Nesi et al., 2018a,b).

These platform features likely influence mental health through core psychological processes, particularly affecting regulation and self-concept development. Adolescents are especially sensitive to social feedback, and social media interactions may shape emotional experiences through interpersonal processes that can either attenuate or amplify negative affect (Lakey and Orehek, 2011). For example, inconsistent or negative feedback may contribute to emotion dysregulation, which has been linked to problematic social media use and internalizing symptoms (Rogier et al., 2024). In parallel, frequent, quantifiable feedback on social media may influence young peoples' developing self-concept and sense of belonging, with potential downstream implications for depression, anxiety, and loneliness.

These processes may be particularly relevant for teens and emerging adults from marginalized backgrounds. Notably, SM use shows differences across groups of young people, with Black, Hispanic, lower household income, and LGBTQ+ adolescents and emerging adults using SM significantly more and qualitatively differently than their White, higher income, and non-LGBTQ peers (Bacaj et al., 2025; Faverio and Anderson, 2025). Likewise, prior work suggests that some rural young people may be vulnerable to internalizing symptoms (i.e., loneliness) because of greater social isolation, fewer accessible services and organized opportunities, and barriers to maintaining supportive connections, although the literature also emphasizes that rural communities are heterogeneous and may offer important protective social ties (Morales et al., 2020; Fenton et al., 2022). Due to isolating experiences in their physical communities, these groups frequently turn to social media to seek the support and connection they cannot find in offline interactions (Escobar-Viera et al., 2018, 2019). Unfortunately, the research focus only on harm has led to less emphasis on studying measurable positive effects of SM use on youth mental health (Ansari et al., 2024).

Meyer's stress theory and overlapping social categories (such as race and ethnicity) can help contextualize the protective and risk factors driving youth mental health disparities. Drawing from the social ecological model (McLeroy et al., 1988), Meyer's stress theory (Meyer, 2003) posits that exposure to “distal” stressors, such as discrimination, coupled with internal processes, such as anticipated rejection, may lead to feeling isolated and negatively impact mental health (Meyer, 2007). Moreover, other social categories, such as race and ethnicity, overlap to impact individuals' mental health, reflecting prevailing sociocultural norms (Crenshaw, 1991; Sokoloff and Dupont, 2005).

Meyer's theory also highlights that social support is a protective factor for mental health, and there is evidence that social support can reduce symptoms of depression, anxiety, and loneliness (Kieling et al., 2011; Gariepy et al., 2016; Rueger et al., 2016; Liu et al., 2021). Perceived emotional support is a form of social support with a strong protective effect on mental health during mid- and late-adolescence, and emerging adulthood (Chu et al., 2010; Yao et al., 2015), potentially through individuals regulating their affect and actions via shared conversations and activities (Lakey and Orehek, 2011). Yet, less than 60% of U.S. young people report they receive the emotional support they need (Zablotsky et al., 2024). Social support processes may be especially salient during adolescence and emerging adulthood, a developmental period marked by increased reliance on peers for validation and belonging. Both the availability and perceived quality of support are critical, with perceived emotional support showing particularly robust associations with internalizing symptoms (Rueger et al., 2016).

One potential protective form of SM experience is that of SM-based emotional support, which refers to expressions of care, empathy, reassurance, validation, and feeling understood that are communicated via SM interactions (Shensa et al., 2020; Munzer et al., 2026). In contrast, online social support is a broader construct encompassing online environments more generally, including emotional, informational, instrumental, and appraisal (Nick et al., 2018). SM-based emotional support represents a specific subset of online social support, distinguished by its platform context and its primary focus on affective exchanges rather than informational or instrumental interactions. SM use is the dominant online activity among adolescents and emerging adults, making it the most ecologically relevant context, and the affective quality of SM interactions may have distinct pathways to mental health outcomes compared to informational or instrumental support more broadly. It is possible that affordances of SM platforms might uniquely enable emotional support processes that affect regulation and perceived social standing, such as asynchronous and low-threshold signals of care (e.g., messages, comments, and reactions), thus increasing the perception of availability of emotional support (Lakey and Orehek, 2011; Rueger et al., 2016). Yet, early evidence on SM-based support and mental health is mixed, with one study's findings showing lower loneliness and higher well-being with small-to moderate effects, especially in youth experiencing discrimination (Best et al., 2014), while another study in young adults found a negative association with depressive symptoms (Shensa et al., 2020). Finally, one meta-analysis found that online social support might increase perceived overall social support and life satisfaction, and reduce loneliness and anxiety, but it recommended longitudinal studies to confirm this effect (Zhou and Cheng, 2022).

Given that many teens and emerging adults expand their social networks and seek support on SM, it is important to examine whether SM-based emotional support could have a protective effect on young people's common mental health symptoms of depression, anxiety, and loneliness over time. These processes may be particularly consequential during adolescence, when identity formation and sensitivity to peer evaluation are heightened. To the best of our knowledge, there are no prior longitudinal studies to examine this question in young people. Therefore, the objective of this study was to examine the associations between perceived emotional support from SM interactions and internalizing mental health symptoms (i.e., depression, anxiety, and loneliness) over time, adjusting for baseline differences in a rich set of demographic and psychosocial characteristics, in a national sample of adolescents and emerging adults from minority backgrounds recruited online. Based on the extant literature, we hypothesized that SM-based emotional support would negatively predict self-reported (H1) depressive symptoms, (H2) anxiety symptoms, and (H3) loneliness over time.

2 Materials and methods

2.1 Participants and procedures

We conducted a longitudinal observational study in which each participant completed online questionnaires three times over a period of 10 months. Wave 1 was completed from November 2023 to March 2024, Wave 2 between April and July 2024, and Wave 3 between August and November 2024. To allow changes in outcomes that are somewhat stable over time (e.g., depression, loneliness), each participant completed the three surveys 3–4 months apart from the other. We recruited participants through social media advertisements posted nationally on Facebook, Instagram, and TikTok. The ads directed interested individuals to a study website, which had more information about the study and a link to a web-based eligibility screener. Individuals were eligible to participate if they were between 13 and 21 years old, resided in the United States, and identified as at least one of the following groups: racial minority (American Indian/Alaskan Native, Asian, Black, Native Hawaiian/other Pacific Islander, Multiracial) or ethnicity minority (Latinx/Hispanic/Latino/a), or sexual or gender minority (Gender minority: Transgender, Non-binary; Sexual minority: Lesbian, Gay, Bisexual, Pansexual, Asexual, Queer, Same-gender loving) or rural living. We included young people living in rural areas as potentially marginalized in relation to internalizing symptoms because marginalization in this context refers to place-based exclusion from resources and opportunities, not to an intrinsic characteristic of rural identity (Graves et al., 2024). To determine rurality, individuals were asked to provide the zip code where they reside, and we verified zip codes by checking if they were designated as rural zip codes by the Federal Office of Rural Health Policy (as of August 2023).

We received a total of 3,965 responses to the screener. To identify fraudulent responses to the screener, we used a stringent set of criteria including completing the screener in under 30 seconds, completing it within three minutes of another response, accessing the screener without clicking on the study website, using an email address already deemed fraudulent in one of our previous studies, and duplicate IP addresses, coordinates, and contact information. Individuals whose responses met any of these criteria were not further considered to participate in the study. Individuals whose responses were deemed legitimate were invited to a “meet and greet” video or phone call, and upon confirming eligibility were emailed an invitation and a link to participate in the study. Using these procedures, 3,276 responses to the screener (82.7%) were considered fraudulent, leaving 689 invited to participate the study. Eligible individuals provided assent or informed consent (for those under or over 18 years of age, respectively) before proceeding with the survey. Participants then completed a web-based survey with a battery of questions about mental health, social media experiences, psychosocial dimensions, and demographic characteristics. We obtained a waiver for requiring parental consent to protect the privacy of LGBTQ+ participants under 18 years of age from having to disclose their identity to their parents or caregivers if they did not feel ready to. Participants were compensated with eGift cards after completing each wave of data collection. All recruitment and data collection procedures were approved by the University of Pittsburgh Institutional Review Board.

2.2 Measures

Responses on social media use, social support, and mental health outcomes were collected at baseline, 16–20 weeks (4–5 months) and 32–40 weeks (8–10 months) after baseline for a total of three time points per participant. At baseline, we collected responses on sociodemographic characteristics and several other psychosocial dimensions. Table 1 summarizes the study population characteristics. We analyzed raw scores for all measures in order to preserve the original metric of the instrument and allow direct interpretation of within-person change over time (Moeller, 2015).

Table 1. Participant characteristics (n = 689).

^aHighest grade and SES were used as continuous covariates in the subsequent analysis.

^bThe number of social media platforms used is the total number of social media platforms each participant uses with a minimum frequency of less than weekly up to several times per day. The platforms included TikTok, Instagram, Twitter, YouTube, WhatsApp, Reddit, Facebook, Tumblr, Snapchat, Pinterest, and Discord.

2.2.1 Internalizing mental health symptoms (dependent variables)

Depressive symptoms were measured with the 9-item Patient Health Questionnaire (PHQ-9) (Kroenke et al., 2001). Items assessed depressive symptoms over the last 2 weeks and were scored using a 4-point Likert-type scale ranging from 0 to 3 (“not at all” to “nearly every day”). Internal consistency is high (α = 0.85–0.89). We summed responses across the nine items resulting in a continuous scale from 0 to 27.

Anxiety symptoms were measured with the 7-item Generalized Anxiety Disorder Scale (GAD-7) (Spitzer et al., 2006). Items assessed anxiety symptoms over the last 2 weeks and were scored using a 4-point Likert-type scale ranging from 0 to 3 (“not at all” to “nearly every day”). Internal consistency is excellent (α = 0.92). We summed responses across the seven items resulting in a continuous scale from 0 to 21.

Loneliness was measured with the 3-item UCLA Loneliness Scale (Hughes et al., 2004), which has been used in adolescent samples in prior studies (Ip et al., 2024; Bo et al., 2025). Items assessed three dimensions of loneliness: relational connectedness (“How often do you feel you lack companionship?”), social isolation (“How often do you feel left out?”), and self-perceived isolation (“How often do you feel isolated from others?”). While the scale yields a single composite score rather than separate sub-scales, the three items tap into both the close relationships and broader social belonging aspects of loneliness captured in the longer 20-item UCLA scale. Each item had a 3-point Likert-type scale ranging from 1 to 3 (“hardly ever” to “often”). Internal consistency is very good (α = 0.72 to 0.87). We summed responses across the items resulting in a continuous scale from 3 to 9.

2.2.2 Social media-based emotional support (independent variable)

SM-based Emotional Support was measured with a modified version of the 4-item Patient-Reported Outcomes Measurement Information System (PROMIS) Emotional Support domain (Carle et al., 2015) that have been used in prior research (Shensa et al., 2020). Items assessed perception of availability of support on social media using a 5-point Likert-type scale ranging from 1 to 5 (“never” to “always”). We modified each statement by adding “on social media” to contextualize the environment to which the statement referred to. Example items included statements such as, “I have people to talk with on social media when I have a bad day.” We summed responses across the four items resulting in a continuous scale from 1 to 20. While SM-based emotional support showed a slight decline over time, as indicated in Table 1, the change was not statistically significant.

Because the SM-based emotional support questions were a modification of in-person validated scales and were previously used with young adult participants, we evaluated the factor structure of the SM-based Emotional Support measure in youth participants via confirmatory factor analysis (CFA) using Mplus (version 8.10). The 1-factor solution was acceptable: Comparative Fit Index (CFI) = 0.996, Tucker-Lewis Index (TLI) = 0.989, Root Mean Square Error of Approximation (RMSEA) = 0.086, 90% confidence interval (CI) = 0.135, Standardized Root Mean Square Residual (SRMR) = 0.099. The Cronbach's alpha for SM-based emotional support was 0.946 with (95% CI: 0.939, 0.954). While CFI, TLI, and SRMR were within acceptable limits, RMSEA fit was not. However, prior work (Kenny et al., 2015) showed that RMSEA can easily reject a model when the model has small degrees of freedom. Our model has only 2 degrees of freedom, suggesting the RMSEA can be misleading in the case of SM-based emotional support.

2.2.3 Covariates

Rich baseline covariates, including the number of social media platforms individuals use, social support from peers, everyday discrimination, and demographic characteristics were collected. We included these factors because they may be associated with different measures of SM experiences such SM-based support and mental health outcomes (Primack et al., 2018; Escobar-Viera et al., 2020; Shensa et al., 2020) (see Table 1 for these covariates).

2.2.4 Analytic approach

All analyses were conducted in R, version 4.4.2 (R Core Team, 2024). We used the R package lmerTest, version 3.1.3 (Kuznetsova et al., 2017) to estimate longitudinal associations between each of the three mental health outcome variables and SM-based emotional support. To answer our research questions, we fitted mixed-effects models that assigned time to Level 1 and individuals to Level 2. The intraclass correlation coefficients for depression, anxiety, and loneliness were respectively 0.67, 0.65, and 0.67, indicating that most of the variability in mental health outcomes lies between individuals. The estimation model is shown below.

Level 1 (Within-individual):

Mentalit=π0i+π1(Waveit)+π2(SMSupportit) + π3(Waveit×SMSupportit)+eit, eit~N(0,σ2)

Level 2 (Between-individual)

π0i=β00+Xβ01+γ0i, γ0i~N(0,σ2) π1=β10π2=β20π3=β30

Note. Xβ_**01** is a vector of covariates listed in Table 1. Dummy indicators were included in the model for categorical variables. e_*it* and γ_0*i* are assumed to be normally distributed.

At Level 1, wave indicator, time-varying SM-based emotional support, and the interaction of the two were included as predictors. The wave indicator was centered at Wave 1, ranging from 0 to 2. The interaction term was included to allow the effect of SM-based emotional support on mental health outcomes to vary over time. At Level 2, baseline covariates in Table 1 were included in the model. We included a random intercept to account for the difference in the average mental health levels between individuals. A total of 689 individuals completed the Wave 1 survey, and missingness in Wave 1 was small. One hundred sixty-nine individuals did not participate in Wave 2, while 122 did not participate in Wave 3. Seventy-nine people did not participate in either Wave 2 or Wave 3 (see the missingness information in Table 1). In the main analysis, we fitted the above model using all available observations, retaining individuals with partial data and relying on the Missing at Random (MAR) assumption (Baraldi and Enders, 2010). Baseline depression, the number of social media platforms used, and living with family were associated with missingness at Wave 2, whereas baseline age and polyamorous relationship status were associated with missingness at Wave 3. To make the MAR assumption more plausible, we included these baseline covariates associated with attrition, along with theoretically relevant predictors of the outcome in the longitudinal model.

3 Results

Seventy-six-point nine percent of our sample was female birth sex, and just over a half of them were White, non-Hispanic (52.2%). The mean age was 19 (standard deviation = 1.52, range 13–22). A total of 323/688 (46.9%) identified as Lesbian, Gay, and Bisexual (LGB), while 158/688 (23.0%) identified their sexual orientation as straight (heterosexual). The characteristics of the sample are presented in Table 1. The average scores for mental health outcomes (i.e., depression, anxiety, and loneliness) tended to be lower in the latter waves. Appendix 1 presents the longitudinal associations between SM-based emotional support and mental health outcomes before adjusting for covariates.

Table 2 presents the results of our mixed-effects linear models of longitudinal associations between mental health outcomes and SM-based emotional support, adjusting for all baseline covariates (the full model with estimates for all the covariates is presented in Appendix 2). In our sample, neither depression nor anxiety scores significantly changed over time, but loneliness significantly decreased across all participants by 0.26 points per wave, on average. Confirming our hypotheses, depression, anxiety, and loneliness were all negatively and significantly associated with SM-based emotional support at Wave 1. In other words, for every 1-unit increase in perceived SM-based emotional support, depression, anxiety, and loneliness scores decreased by 0.09, 0.11, and 0.07 points at Wave 1, respectively. The non-significant interaction terms between wave and SM-based emotional support suggested that the negative associations between mental health outcomes and social media emotional support were relatively stable over time. Appendix 2 presents the longitudinal associations between SM-based emotional support and mental health outcomes after adjusting for covariates, along with estimates for all covariates.

Table 2. Longitudinal associations between mental health symptoms and SM-based emotional support adjusting for covariates.

^**p < 0.01,

^*p < 0.05.

All covariates listed in Table 1 were controlled for in the model. Dummy variables were included for categorical variables. Baseline covariates included the number of social media platforms used, social support from peers, everyday discrimination, negative life events, age, race indicators (White, Black, Asian, Hispanic, and Multiracial), education level, sex at birth (female and unsure), sexual orientation (lesbian, gay, bisexual, pansexual, asexual, queer, and straight), gender (female, male, trans, and non-binary), socioeconomic status, relationship (polyamorous, legal, and no relationship), living situation (alone, with family, and with peers), and rural (see Appendix 2 for the full results). The wave indicator was centered at wave 1. For instance, when wave = 0, the results show the associations at wave 1.

3.1 Robustness check

We conducted a sensitivity analysis by running the same models for individuals without missing cases in the subsequent waves (Waves 2 and 3). The results were consistent with the original findings (see Appendix 3). Because the patterns may vary by developmental stage (i.e., adolescence vs. emerging adult), we tested whether the longitudinal associations between mental health and SM emotional support differed by participants' age groups. We did not find significant age moderating effects (see Appendix 4).

4 Discussion

4.1 Main findings

Emotional support is essential for completing key developmental tasks during adolescence and young adulthood, including identity formation and social reorientation (Rageliene, 2016; Branje, 2022). Family-, peer-, and school-based support are widely considered prominent sources of emotional support for young people, and their protective effect on mental health has been well-documented (Butler et al., 2022). Prior research on the role of SM-based emotional support in mental health has yielded mixed findings, though most studies relied on cross-sectional data. We sought to examine the longitudinal association between SM-based emotional support and internalizing symptoms in young people. In a national sample of 689 U.S. teens and emerging adults, results confirmed our hypotheses: while perceived availability of emotional support on social media was associated with significantly lower depression, anxiety, and loneliness symptoms at Wave 1, the rate of change over time was only significant for loneliness, with all associations remaining stable across the study period.

The main findings of our longitudinal analysis complement and expand prior cross-sectional studies suggesting the potential of perceived online social support as a protective factor for young people's mental health (Zhou and Cheng, 2022; Munzer et al., 2026). These findings also complement a growing literature linking problematic or dysregulated social media use to internalizing symptoms in adolescents (Muzi et al., 2021), highlighting that social media may confer both risk and resilience, depending on context and how it is used and experienced. This could be important in the context of young people living in rural areas, where some youth might lack resources and opportunities to build community and socialize, making SM an accessible resource to bridge those gaps (Fenton et al., 2022). Unfortunately, the literature also highlights that digital exclusion based on geographic location is a reality for some rural young people, and these individuals may not gain benefits from SM either (Graves et al., 2024). While the associations between our main independent variable (SM-based emotional support) and each of the three main outcome variables (i.e., depression, anxiety, and loneliness symptoms) were significant at Wave 1, the main effects of SM-based emotional support were small on depression and loneliness, to moderate on anxiety symptoms. Notably these associations remained after adjusting for in-person peer support, number of SM platforms used, and a comprehensive set of sociodemographic characteristics. In addition, the non-significant interactions between main independent variable and wave (i.e., time point) in each of our three main models indicate that the associations between SM-based emotional support and depression, anxiety, and loneliness symptoms remained relatively stable over time.

The small to moderate coefficients of SM-based emotional support on mental health outcomes measured in this study raise the question of whether SM-based emotional support holds the potential of being a protective factor for mental health. There are three paths in which this could be the case. First, it is possible that SM does not need to deliver intensive help to be protective; prior research demonstrated that perceived support is strongly associated with lower depression and anxiety symptoms in teens and emerging adults (Lakey and Orehek, 2011; Rueger et al., 2016). SM platforms have unique affordances that enable asynchronous, low-threshold signals of care, such as messages, comments, and reactions, which can strengthen perceived availability of support that may be sufficient to alter individuals' stress appraisals. These findings are also consistent with mechanisms of affect regulation, whereby even low-intensity supportive interactions may facilitate recovery from negative emotional states (Lakey and Orehek, 2011).

Second, SM-based emotional support might function as a buffer during acute periods of psychological distress (e.g., social conflict, identity stressors), a context in which prior work suggests that both beneficial and problematic forms of social media use may have amplified associations with internalizing symptoms (Muzi et al., 2021). The stress-buffering effect of emotional support on internalizing symptoms is robust and well-studied (Cohen and Wills, 1985; Bekiros et al., 2022); longitudinal studies have shown negative associations between social support and depressive symptoms following episodes of stress (Tian et al., 2025). In support of this path, research has shown that online environments such as SM have similar buffering patterns; teens reporting higher online emotional support show reductions in the association between stress and depression symptoms (Frison and Eggermont, 2015, 2020). Finally, it is also possible that SM-based emotional support may be especially protective when it provides identity-affirming belonging unavailable in offline contexts, thereby reinforcing self-concept and perceived social standing. Adolescence and emerging adulthood are periods of high sensitivity to social inclusion and rejection, and among young people from marginalized backgrounds (e.g., racial or ethnic minority, LGBTQ+), supportive online interactions may facilitate identity exploration (Uhls et al., 2017). Additionally, online support is associated with lower depression and increased well-being, often compensating for hard-to-find social support in their physical environments (McConnell et al., 2017; Escobar-Viera et al., 2018). Regardless of the potential path through which SM-based emotional support might confer protection against internalizing symptoms, evidence shows that these effects are not universal and are more likely to be realized when support is reciprocal, validating, within a trusted or semi-private space, and when young people are experiencing stress or marginalization (Nesi et al., 2018a,b; Pellicane et al., 2021). Future research should focus on leveraging more data-intensive methods (e.g., ecological momentary assessment, machine learning, other forms of artificial intelligence), that maximize ecological validity, remove participant burden, and can ensure generalizability for in-depth examinations of what types of SM content (e.g., direct messages, images, posts, reactions) and interactions convey support.

In this study, we found a unique pattern in relation to loneliness scores, as these decreased a quarter of a point per wave. On the one hand, this reduction in self-reported loneliness may reflect some developmental processes, such as improved social skills or expanded peer networks, or consistent perception of emotional support, which could be facilitated via SM interactions (Smith and Alheneidi, 2023; Käcko et al., 2024). On the other hand, it could also reflect other factors that remained unmeasured in our models.

Lastly, we had a high proportion (82.7%) of responses to the initial screening questionnaire disseminated via social media that were classified as fraudulent and excluded before enrollment into the longitudinal study. Although substantial, this proportion is consistent with growing methodological literature showing that SM recruitment can attract large numbers (between 60%−95%) of invalid responses, including bots, duplicate entries, and ineligible respondents attempting to obtain study incentives (Arundell et al., 2016; Pozzar et al., 2020; Cho et al., 2025). For this reason, we treated the screening questionnaire not as part of the analytic sample, but as a gatekeeping step designed to protect data integrity.

4.2 Limitations

Through our study, we gained new insights on the protective effect of social media-based emotional support on internalizing symptoms in a national sample of U.S. youth, with overrepresentation of marginalized youth. Nevertheless, our work is not without limitations to consider. First, the measure used to assess SM-based emotional support was adapted from an in-person emotional support measure originally developed and tested with adults (Carle et al., 2015). While the adapted measure had been validated in young adults before (Shensa et al., 2020), it is still possible that the measure did not capture dimensions of SM-based emotional support important to adolescents. We addressed this by conducting a CFA. Though acceptable, CFA results suggest there might be room for improving the validity of the SM-based emotional support measure as it is used with adolescents.

Second, recruitment advertisements on only three SM platforms might have introduced selection bias. While SM use is ubiquitous in teens and emerging adults and we posted our ads on popular SM among young people (e.g., Instagram and Tiktok), individuals who did not use these platforms did not have a chance to be selected into the study. Third, when we modified the PROMIS emotional support items to adapt them to social media, we did not specify to participants which types of SM sites/apps they should consider. On the one hand, this allowed participants to widely think about the platforms from which they derive the most support, regardless of whether the platform is more visual-focused (e.g., TikTok, Instagram) or text-based (e.g., Facebook, Tumblr, Discord). On the other hand, we might have missed potential differences in perceived emotional support among frequent users of one or another type of platforms. Fourth, in our analysis, we could not adjust our models for time spent on SM. While the results are mixed and self-report of this measure is often inaccurate (Sewall et al., 2020), most of the research on SM and youth mental health has been conducted using time spent on SM as the main independent predictor. To address this, we used a proxy measure of volume of SM use and adjusted our models for the self-reported number of SM platforms used.

Fifth, we adjusted for a rich set of sociodemographic factors, but a few unmeasured variables remain that may explain our findings. For example, personality traits, seasonality, baseline mental health trajectories, and offline social network quality could influence our primary outcomes. To attenuate some of these limitations, we did adjust for adverse experiences and offline peer-support. However, future research should consider assessing these factors, denser repeated assessments, or calendar-time adjustment. Sixth, there is a possibility of reverse causality between the variables. It is possible that youth who felt less depressed, anxious, or lonely elicited more SM-based emotional support. To disentangle the role of baseline mental health trajectories and reverse causality, future longitudinal studies should consider designs that allow examining bidirectional associations.

Finally, the magnitude of the effects between SM-based emotional support, depression, and loneliness symptoms were small, which could be due to increased rates of false positives in large samples (Ferguson, 2009). Nevertheless, prior research has shown that relatively small changes in average scores of mental health outcomes can indicate larger shifts when scaled up to population level (Carey et al., 2023), and this effect could be particularly significant for youth from marginalized backgrounds (Nesi et al., 2018a,b; Pellicane et al., 2021).

4.3 Conclusion

This study sought to determine whether perceived emotional support from SM could reduce symptoms of depression, anxiety, and loneliness in youth. Our main findings indicate that in a national sample of U.S. youth, SM-based emotional support can confer a small to moderate protective effect for internalizing symptoms, even in the presence of in-person forms of emotional support, and this effect remained stable over the study period. These findings suggest that increasing the perception of availability of emotional support on social media in teens may be complementary and additive to more traditional types of social support, especially for adolescents who experience marginalization or discrimination.

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/fdpys.2026.1811938/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Sarah Domoff, University at Albany, United States

Reviewed by: Stefania Muzi, University of Genoa, Italy

Aubrey Borgen, Square One Specialists in Child and Adolescent Development, United States