We’ve known for a decade that clinical-level depression, self-harm, suicidal thoughts, and suicide rose among adolescents after 2011.
We also know that the more time a teen spends using social media the more likely it is he or she is unhappy, low in life satisfaction, or high in symptoms of depression.
There’s a bit of a difference, though: The studies of change over time have shown increases in mental health issues that require treatment, while the studies among individuals mostly focus on low psychological well-being or symptoms of depression measured by questionnaires. This difference has led some to argue that there is no evidence of a link between social media use and mental health disorders that require treatment.
I’ve always been skeptical of this argument given the strong links between self-report well-being or symptom measures and mental health disorders; the self-report measures correlate strongly with clinical diagnoses and are often used to screen for mental health disorders. It’s pretty unusual to find someone with major depression who says they are happy and doing well.
Still, having data showing a link with mental health disorders would be even better. That way, we’d know for sure that social media use was associated with actual mental health disorders – the type that require treatment – not just with unhappiness, low life satisfaction, or elevated symptoms of depression.
My colleagues and I have done a few of these studies. We found a link between social media use and self-harm behaviors in a large sample of UK adolescents, though self-harm was measured via self-report. We also found that digital media use was linked to depression, self-harm, and suicidal thoughts according to clinical criteria among 9- to 10-year-olds in the U.S. in the ABCD dataset. However, social media use was low in this very young population.
Ideally, we’d want a large sample of adolescents with clinically diagnostic measures of mental health disorders as well as a solid measure of social media use. A study that came out last year (Fassi et al., 2025) does just that …