A new social media anxiety study 2025 published in Nature Human Behaviour finds that adolescents with diagnosed mental health conditions use platforms like Instagram, TikTok and Snapchat in measurably different ways from peers without such conditions, a result that quietly reshapes the long-running debate about whether platforms cause anxiety or simply attract the already-anxious. Social media anxiety is the cluster of worry, comparison and physiological arousal that builds in people who use these platforms in ways that begin to feel out of their control, and the new paper argues the people most exposed to it look different from the average teenager before the scrolling even begins. Our Anxiety topic hub covers the broader clinical picture; this brief sticks to what one study moved.
What the study did
The paper, led by Luisa Fassi with co-authors Amy M. Ferguson, Andrew K. Przybylski, Tamsin J. Ford and Amy Orben, draws on a UK adolescent sample large enough to compare two groups directly: young people who already meet criteria for at least one diagnosable mental health condition, and young people who do not. The authors examined patterns of platform use rather than simply totalling daily minutes, including who participants follow, how reactive they are to feedback, how often they compare themselves with others, and how strongly they report feeling unable to stop.
That design matters. Most prior work in this field has chased a single exposure number, usually self-reported hours per day, and tried to predict mental health outcomes from it. The Fassi paper instead asks what the texture of use looks like, and whether that texture differs by clinical status. It is the kind of question that survey-only studies struggle to answer cleanly, and it required the authors to layer multiple measures of use against carefully collected mental health diagnoses.
What the study found
The headline result is straightforward: across nearly every measure of use the team examined, adolescents with diagnosed mental health conditions reported a different relationship to social media than their peers, not just a more intense one.
1. More time, but that was the smallest part of it. Yes, adolescents with diagnosed conditions reported more total time on social platforms. But total time was the least interesting finding. If headline minutes were the whole story, the gap would have been narrow.
2. More social comparison and feedback sensitivity. The clinical-condition group reported sharper upward social comparison, feeling worse after seeing peers’ posts, and more emotional reactivity to likes, comments and the absence of either. This is the kind of pattern the 2024 Journal of Adolescent Health systematic review on social media and adolescent anxiety has flagged as more tightly tied to symptoms than raw exposure.
3. Lower satisfaction with the time spent. Adolescents with mental health conditions were more likely to describe their social media time as something they wanted to do less of but could not. That subjective loss of control is exactly what problematic-use scales, the ones our biggest-findings roundup treats as the tightest predictor of anxiety, try to capture.
4. Different network composition. Beyond behaviour, the shape of who participants were connected to differed. Reported network features, the mix of close friends, strangers, public figures and parasocial accounts, varied with clinical status in ways the authors interpret as relevant to exposure to comparison content.
How it fits the 2025 evidence pool
This is not a one-off finding. The 2025 Behavioral Sciences meta-analysis pooling 24 primary studies on social media and adolescent mental disorders found an aggregate effect of around r ≈ 0.22, meaningful, but small enough that it cannot plausibly be explained by a single direction of causation. If platforms simply caused mental illness in proportion to exposure, the pooled effect would be larger and would not depend so heavily on whether researchers measured “problematic” use or raw screen time.
The Fassi paper helps explain why those pooled effects look the way they do. If adolescents with pre-existing anxiety, depression or related conditions arrive at social media already primed for comparison, already more sensitive to feedback, already more likely to lose track of time, then any aggregate study that lumps them in with their peers will look like a mild dose-response, when the underlying pattern is closer to two different relationships with the same tool. Our contrarian read of the 2025 evidence pulls on this thread further, and our overdiagnosis-debate piece sits squarely with the implication.
That does not let platforms off the hook. The same logic implies a real feedback loop: more anxious teens are pulled toward the most reactive features of the feed, those features then deepen the pattern, and the loop tightens. That bidirectional reading is consistent with the broader 2024–2025 longitudinal literature, including the Journal of Adolescent Health systematic review and the 2025 Behavioral Sciences meta-analysis.
Limitations worth naming
Three limitations should travel with this finding.
First, the study is observational. It cannot tell us whether the different patterns of use caused anxiety in the clinical-condition group, resulted from their pre-existing symptoms, or, most likely, both. The authors are explicit about this. Headlines summarising the paper as “social media causes mental illness in anxious teens” misstate what an observational design can claim.
Second, the sample is UK-based and skews toward English-speaking adolescents in particular socioeconomic strata. The 2024 Journal of Adolescent Health systematic review has cautioned that generalising findings across very different platform ecosystems, for example, comparing UK Instagram use with TikTok use in another regulatory environment, is harder than aggregate effect-size language suggests. The gendered differences other recent work has flagged are not the focus here either; readers interested in that split should see our Gen Z gender breakdown.
Third, self-report. Almost every variable in the paper depends on adolescents accurately describing their own use, their own feelings about it, and their own diagnostic status. Self-report drift is a known issue in this literature, and there is some evidence that mental-health status itself shifts how people recall and report screen time. The authors compensate with careful design, but the underlying constraint stays.
What it means for readers
For parents, clinicians and curious adults, the practical takeaway is to stop asking “is social media bad for teens?” as if it had one answer. The Fassi study suggests two answers, depending on which teen you are asking about. For the average adolescent without a mental health condition, the picture remains one of small, mostly manageable effects. For adolescents already coping with anxiety or related conditions, social media looks more like a setting that amplifies what is already there, both because they spend more time on it, and because the features of the platform pull harder on the cognitive patterns that drive their symptoms.
The clinical implication is that screen-time blanket rules are a poor substitute for paying attention to how a young person is using these tools. Parents and clinicians who want a useful conversation should look at compulsion, comparison and reactivity rather than minutes. Adults running households where teen mental health is in play may find more recognition than they expect in our piece on mothers and teens and the anxiety they share.
The Fassi paper does not settle the social media debate. It shifts the question to one the field has been slow to ask: who is using this, how, and why, and how do we tell those stories apart before we draw policy conclusions from a single average.