Screen time is a poor variable. It aggregates video calls with grandparents, homework, watching films, messaging friends, and scrolling through images of people who look better than you, and treats them as the same thing.
Unsurprisingly, it predicts wellbeing weakly. Studies using total-use measures typically find associations that are statistically detectable and small.
Studies that separate types of use find considerably more.
Active versus passive use
The most-studied distinction is between active use — posting, messaging, direct interaction — and passive use — scrolling and viewing without interacting.
Several studies have found that passive use is associated with declines in wellbeing while active use is associated with neutral or positive effects. Some experimental work has supported the direction.
The distinction has been challenged and the evidence is not uniform. But it is a considerably better predictor than total time, which is the main point.
The proposed mechanism is comparison. Passive consumption of others' curated presentations provides a stream of upward comparisons with no interaction to contextualise them.
Appearance comparison specifically
Within passive use, appearance-focused content shows the strongest associations with body dissatisfaction and mood, particularly among adolescent girls.
Studies distinguishing content types find that image-based platforms with appearance-focused content show stronger associations than text-based platforms or platforms centred on interests.
Experimental studies exposing participants to appearance-focused content versus neutral content have found short-term effects on body satisfaction and mood, with effects larger among participants with higher baseline appearance comparison tendency.
That moderation is important. The same content affects different people very differently, which is why average effects understate the harm to those most affected.
The comparison target matters more than the medium
Research on social comparison predates social media by decades and identified the relevant variables long ago.
Upward comparison with a similar and attainable target can be motivating. Upward comparison with a target perceived as unattainable is demoralising. Comparison with people close to you affects self-evaluation more than comparison with distant figures.
The last point explains something specific about why social media differs from traditional media. Comparison with a professional model was always available in magazines and was mediated by obvious distance. Comparison with someone in your year group, whose presentation is curated but appears ordinary, is a comparison with a similar target, and similar targets carry more weight.
This suggests that the problematic content is not necessarily the most extreme. It is the plausible.
Displacement
A second mechanism operates independently of comparison and receives less attention.
Time is finite. Time online displaces sleep, physical activity, in-person interaction and unstructured time, and all four have their own well-evidenced relationships with adolescent wellbeing.
Sleep displacement is the best supported. Device use before sleep and overnight notification checking both associate with reduced sleep duration and quality, and the causal evidence linking sleep to mood is stronger than almost anything else in this area.
An intervention that improves sleep may produce more benefit than one targeting content, and it is considerably easier to implement.
What the intervention research shows
Studies asking participants to abstain from social media entirely have produced mixed results, with some finding wellbeing improvements and others finding little or negative effects, plausibly because abstention also removes genuine social connection.
Studies reducing rather than eliminating use, or targeting specific types of use, have generally produced more consistent results.
Interventions targeting comparison directly — through media literacy, through unfollowing appearance-focused accounts, through prompts about curation — have shown promise in small studies.
The pattern suggests that blanket restriction is a blunt instrument and that targeted change is more effective, which is also more practical for anyone who has attempted the former with an adolescent.
What this implies practically
Audit content rather than time. The useful question is what an hour consists of, not how many there are.
Unfollowing is the single most effective available action and is under-used. Feeds are curated by the user; a feed producing consistent appearance comparison can be changed in twenty minutes.
Protect sleep specifically. Devices out of the bedroom has better evidence behind it than most content-focused interventions.
Distinguish connection from consumption. Messaging friends is not the same activity as scrolling, and treating them identically in a household rule produces resistance for good reason.
And attend to the individual rather than the average. The heterogeneity finding is the most consistent result in this literature: an adolescent with high appearance concern consuming appearance content is in a different situation from one messaging friends about a game, and a policy that cannot tell them apart will be both ineffective and resented.