A study involving 2.8 million children sounds as if it should end an argument. The new evidence on screen use does something more useful: it shows where concern is justified, while exposing how crude the phrase “screen time” has become.
Published on September 28 in JAMA Pediatrics, the research brings together results from 23 earlier meta-analyses spanning 64 countries. More screen exposure was associated with weaker language and academic outcomes and with more emotional, peer, and behavioral problems. Those findings matter. But they do not prove that screens, by themselves, caused the differences—and they do not support treating a video call, a classroom assignment, and late-night algorithmic scrolling as the same experience.
The central question is therefore not whether screens are simply “good” or “bad.” It is what the evidence can actually tell us about risk, direction, and the parts of digital life that may be worth changing.
The signal is broad, but usually small
The new umbrella review synthesized data from 2,817,252 children and adolescents aged 18 or younger. Across the included research, greater screen exposure correlated with lower academic achievement (r = −0.15) and language skills (r = −0.14). It was also associated with more internalizing problems, a category that includes anxious and depressive symptoms (r = 0.10 for general screen use), and more externalizing behavior such as aggression or attention difficulties (r = 0.11).
These are not zero effects, especially when exposure is widespread across a population. Yet most are small correlations. The larger associations reported for emotional problems and peer problems came with wide confidence intervals, meaning the true size is less certain. Screen use was not significantly associated with executive functioning, prosocial behavior, or scores in mathematics, science, and reading, although some of those outcomes had been studied less often.
That uneven pattern is important. If screens produced a single, powerful developmental effect, we would expect more consistent results across domains. Instead, the review suggests a modest risk signal whose size changes by outcome, age, screen type, and behavior measured.
Headlines may also overread the authors’ comparison with other developmental factors. An association being similar in statistical magnitude to one involving socioeconomic status or preterm birth does not make the exposures biologically equivalent. Nor does it establish that removing screens would reverse the observed difference. It tells us how variables moved together in the available studies—not why they did.
A large sample cannot solve the direction problem
An umbrella review sits at the top of an evidence pyramid, but it still inherits the weaknesses of the studies beneath it. Much of the screen-time literature relies on self- or parent-reported estimates, broad categories of use, and cross-sectional snapshots. Families also differ in stress, income, sleep, parenting routines, neighborhood safety, and access to offline activities. Those factors can shape both digital habits and children’s wellbeing.
Direction is especially difficult. A child who is distressed, lonely, or struggling with behavior may turn to a device more often; parents may also offer screens more frequently when family life is strained. In that case, heavier use can be partly an outcome rather than only a cause.
A 2025 meta-analysis of 117 longitudinal studies involving 292,739 children found small associations in both directions: greater screen use predicted later socioemotional problems, and earlier socioemotional problems predicted greater subsequent screen use. Gaming showed stronger links than other forms of screen activity, but even longitudinal evidence is not the same as random assignment. Unmeasured differences can still influence both sides of the relationship.
The same caution applies to a large prospective analysis of 9,538 US children. More screen time at ages nine and ten was associated with slightly higher mental-health symptom scores one and two years later, even after adjustment for baseline symptoms and several demographic factors. The authors described the effects as small and noted that screen time did not capture content or context; inverse causality also remained possible.
Scale makes an estimate more precise. It does not automatically turn an association into a mechanism.
The better question is what the screen is doing
“How many hours?” remains useful, but it is rarely sufficient. An hour of collaborative homework, creative work, or conversation with a distant grandparent has different social and cognitive ingredients from an hour of solitary, compulsive use before sleep. Age matters. So do content, timing, design, and whether a screen displaces movement, rest, reading, or face-to-face contact.
That more contextual approach now appears in the American Academy of Pediatrics’ 2026 policy statement. Rather than offering one universal number for every child, it emphasizes high-quality content, time boundaries that fit family routines, shared use, and protecting sleep, play, physical activity, homework, and family time. It recommends screen-free settings such as meals and bedrooms and avoiding screens during the hour before bed.
This is not evidence that limits are pointless. The new review strengthens the case for paying attention when digital media consistently crowds out essential parts of childhood. But it argues against panic and against using a stopwatch as the only measure of digital health.
The clearest conclusion from 2.8 million children is narrower—and more actionable—than “screens harm kids.” Longer exposure is linked to several less favorable outcomes, usually modestly, while the reasons are mixed and sometimes reciprocal. The most productive response is to examine not only duration, but purpose, pattern, and displacement: what the child is doing, why they are doing it, and what that time has replaced.
