The evidence, in plain English
40+ peer-reviewed studies and meta-analyses, sorted by what they found — on reading, attention, mental health, device programs, and who gets hurt most. Screens earn a place for the few things they genuinely do better. As the everyday default, the evidence runs the other way.
Schools went screen-first fast, and the research has been catching up. The pattern across it is consistent enough to say plainly: for the everyday work of a classroom, analog and human-led methods beat screen-based ones.
Reading comprehension is reliably better on paper than on screens — four separate meta-analyses, 170,000+ readers, and the gap has grown over time. Writing by hand builds more of the brain connectivity tied to memory than typing does. The biggest one-laptop-per-child programs studied at population scale lowered test scores, not raised them.
The harm isn’t spread evenly. The device programs widened achievement gaps, hitting already-disadvantaged kids hardest — which turns the “close the digital divide” pitch on its head. And the AI tools schools adopt carry measurable bias: speech recognition mis-hears Black students, AI-writing detectors falsely flag English learners, proctoring software flags darker-skinned kids more often.
None of this says a screen is never the right tool. It says the burden of proof belongs on the screen. Use one deliberately for the jobs it does better — a simulation you can’t run on paper, an accessibility need, a window onto something far away — and keep the rest of the day built on the things with the strongest evidence behind them: teachers, books, handwriting, and time outside.
The research
Ten domains, sorted by the worry they answer. Each study links to its source — and, where we’ve translated it, to a plain-English version you can actually read.
Brain & cognition
The first MRI study of preschoolers found that kids with screen use above the pediatric guidelines had measurably lower structural integrity in the white-matter tracts that carry language and early literacy — and scored lower on language tests to match. A follow-up tied higher screen use to thinner cortex in regions for attention, memory, and reading. Reading a printed book and watching the same story on a screen don't light up the brain the same way: print produces stronger, more balanced connectivity, and a parent and child reading print together show more synchronized brain activity than they do on a screen.
Preschoolers above screen guidelines showed lower white-matter integrity in language and literacy tracts.
Higher screen use in 3–5-year-olds linked to thinner cortex in attention, memory, and letter-recognition regions.
Language-and-imagery brain networks were strongest during illustrated storybook reading, not animation.
Book-reading time raised brain connectivity; screen time lowered it in the same networks.
Parent–child neural synchrony was greater during printed-book reading than screen reading.
More frequent internet use predicted smaller gains in brain volume and verbal intelligence over time.
Handwriting vs. typing
High-density EEG shows handwriting produces far more widespread brain connectivity — in exactly the regions tied to memory and learning — than typing the same words. The pattern holds in 12-year-olds, not just adults, which puts it squarely in K-12 territory. And students who take notes by hand understand the material better than laptop note-takers, who tend to transcribe verbatim instead of processing in their own words (a direct 2019 replication found the effect smaller and not statistically significant, so we flag it rather than overstate it).
Handwriting produced widespread learning-and-memory brain connectivity; typing did not.
The same handwriting-superior connectivity appeared in 12-year-olds — directly relevant to K-12 policy.
Laptop note-takers scored worse on conceptual questions than longhand note-takers.
Compared keyboarding-instruction techniques for building elementary students’ skills.
Reading: paper vs. screen
Four separate meta-analyses — covering hundreds of studies and well over 170,000 readers — land in the same place: comprehension is reliably better on paper than on screens, especially for informational text and under time pressure. The screen gap has grown, not shrunk, in studies of kids who grew up with screens, and readers on screens are more overconfident — they think they understood more than they did. For young children, poorly designed digital 'enhancements' actively get in the way, while an adult reading a print book alongside a child beats the child reading a digital book alone.
54 studies, 170,000+ readers: paper beat screens for comprehension, and the gap has widened over time.
Confirmed the paper advantage for informational text; screen readers were more overconfident.
A third meta-analysis: comprehension better on paper, with no time savings from screens.
Children's shared reading: digital enhancements impede comprehension; adult-mediated print wins.
Attention & distraction
A two-year study of teens with no baseline ADHD symptoms found that heavy digital-media use predicted developing those symptoms; a review of 28 longitudinal studies points to a reciprocal loop, with problematic use feeding later attention problems partly through lost sleep. In-class laptop multitasking lowers test scores not only for the user but for nearby students who can see the screen. And across OECD countries, students who spend 5–7 hours a day on devices at school scored 49 points lower in math — and device distraction costs the equivalent of roughly three-quarters of a school year.
~2,600 teens: high digital-media use predicted developing ADHD symptoms over two years.
28 longitudinal studies: digital media use contributes to later ADHD symptoms, partly via sleep.
Laptop multitasking lowered test scores for the user AND nearby peers who could see the screen.
Heavy in-school device users scored 49 points lower in math; distraction ≈ ¾ of a school year.
EdTech & 1:1 device programs
When researchers track real device programs at population scale, the academic results are negative. Nationally representative U.S. data found EdTech use in grades K-3 lowered achievement and widened gaps, with the harm concentrated on already-disadvantaged kids. Catalonia's one-laptop-per-child program cut performance in four subjects by 0.20–0.22 standard deviations. Expanding home computers and broadband in North Carolina produced persistent declines in reading and math; a Romanian voucher lottery found kids who won home computers got lower grades. The 'digital divide' framing predicts the opposite of what the strongest evidence shows.
U.S. K-3 EdTech use lowered achievement and widened gaps for disadvantaged students (ECLS-K).
Catalonia's 1:1 laptop program cut scores 0.20–0.22 SD across four subjects, worse for boys.
~1M students: home computer/broadband access produced persistent reading and math declines.
Voucher lottery: kids who won home computers had lower grades in math, English, and Romanian.
Mental health
In a sample of over 40,000 U.S. kids, heavy screen users (7+ hours a day) were more than twice as likely to have been diagnosed with depression or anxiety. A four-year study found that within the same teen, rising social-media and TV use predicted rising depressive symptoms. Two large meta-analyses show the risk climbing with each additional daily hour in a dose-response pattern, and the prospective-only evidence holds: more screen time comes first, more depression risk follows.
40,000+ kids: 7+ hrs/day screen users were 2× as likely to be diagnosed with depression/anxiety.
~3,800 teens, 4 years: within-person rises in social media and TV predicted rising depression.
Problematic social-media use correlated with depression, anxiety, and stress across 18 studies.
26 studies, 55,340 teens: depression risk rose with each added hour of social media (dose-response).
Prospective cohorts only: higher screen time predicted elevated later depression risk.
Sleep, vision & physical health
Children aged 6–19 with a screen device in the bedroom had roughly double the odds of too little sleep, poor sleep quality, and daytime sleepiness — and just having the device, used or not, was harmful. Across the research, 90% of studies tie screen time to worse sleep. More screen and near-work time raises the risk of myopia (nearsightedness), while outdoor time protects against it; higher screen time also tracks with childhood overweight. The sleep harms begin in infancy and toddlerhood, which is why analog and outdoor defaults keep showing up in the recommendations.
Bedtime device access ≈ doubled the odds of inadequate, poor-quality sleep (even access alone).
90% of studies found screen time adversely associated with sleep in school-aged kids and teens.
Systematic review linking digital screen time to elevated myopia risk in children.
19 studies: higher screen exposure linked to elevated myopia risk; outdoor time protective.
Higher screen time significantly associated with childhood overweight/obesity.
Screen time tied to poorer sleep from infancy; activity and outdoor play favorable.
Addictive & persuasive design
Of apps used by 3–5-year-olds, 80% contained manipulative 'dark pattern' design — fake time pressure, parasocial pressure, lures — and nearly every child had at least one, with the manipulation most common in apps used by lower-income kids. Most top-rated 'educational' apps score low against the actual science of learning, so the marketing claims don't hold up. And apps played by preschoolers routinely transmit persistent identifiers to third parties, raising the privacy questions the law was written to prevent.
80% of apps used by 3–5-year-olds had manipulative design; worst in apps used by lower-income kids.
Most top "educational" apps scored low on real science-of-learning criteria.
Preschoolers’ apps frequently sent persistent identifiers to third parties (COPPA-relevant).
Equity & algorithmic bias
Screen-heavy defaults concentrate their harms on the kids they're sold to help. The device programs above widened achievement gaps rather than closing them. The tools themselves are biased: commercial speech recognition makes about twice as many errors on Black speakers, AI-writing detectors falsely flag more than half of essays by non-native English writers, and proctoring software flags darker-skinned students as suspected cheaters more often. And the surveillance these devices import falls hardest on Black, Hispanic, disabled, and LGBTQ+ students — used more for discipline than safety.
All five major speech-recognition systems made ~2× the errors on Black speakers vs. white speakers.
AI-writing detectors misflagged >50% of non-native English essays as AI-generated.
Automated proctoring flagged darker-skinned students as suspected cheaters more often.
Filters disproportionately block LGBTQ+ and race-related content; monitoring drives biased discipline.
Device monitoring used more for discipline than safety; 29% of LGBTQ+ students reported nonconsensual outing.
Parents confirm Black, Hispanic, and LGBTQ+ students bear disproportionate monitoring harm.
Not a fringe read
This isn’t one contrarian study. Major institutions and at least one national government have read the same evidence and changed course.
The 2023 Global Education Monitoring Report calls for caution: technology should never replace in-person, teacher-led instruction, and much of the favorable evidence for classroom tech is funded by the companies selling it.
Sweden's medical university told the government the evidence shows digital tools impair rather than enhance learning. Sweden then reversed mandatory devices in preschools, ended digital learning under age 6, and funded a return to printed textbooks after national reading scores fell.
The AAP's 2026 review of generative AI concludes the risks to children vary by developmental stage and are most acute in early and middle childhood.
A 2025 open-access synthesis by more than 100 researchers — the most complete single picture of the evidence base from birth through adolescence.
Failure to Disrupt (Harvard University Press) synthesizes decades of evidence: scaled learning technologies are used and benefited from disproportionately by already-advantaged learners, so tech-first reforms tend to reproduce inequality.
For your family
Tell us their grade, what the school runs on, and what’s worrying you — and we’ll write the version of this that fits your situation, with the next step that matches what you’re trying to do.
Sign in to personalize this guide with data from families in your school, city, and community
See your own school’s picture.
The five-minute Screens in Schools survey collects what’s actually happening in your kid’s building — devices, AI tools, screen hours — and shows it back to you alongside other families at the same school. Numbers beat vibes when you talk to a teacher or PTA.
Take the Screens in Schools SurveyWhat the evidence supports
Every domain above points back to the same short list. Printed books, which beat screens for comprehension. Handwriting, which beats typing for what kids retain. Small-group and one-on-one human instruction. Time outside, which protects sleep and vision. These are the equity investments with the deepest evidence base — not because they’re old-fashioned, but because they work.
Screens still earn a spot for specific jobs: a simulation you can’t build on paper, an accessibility tool that opens a subject up, a video of a place a kid will never visit. The line isn’t screens-or-nothing. It’s deliberate-or-default — and the default should be human.
If you want to act on it
When a school puts a device or an app in front of your kid, a stack of laws governs it. Here’s the framework — written for any U.S. parent, with D.C. Public Schools as the worked example — and the two requests worth putting in writing. This is a starting map, not legal advice.
Under COPPA’s school-consent rule the legal duty sits with the ed-tech vendor — it owes the school notice of what it collects plus tools to review, delete, and stop collection — and the FTC urges schools to pass that on to parents. That makes these two requests fair to put in writing and keep on the record:
The Children's Online Privacy Protection Act governs data collection from under-13s. A school may consent on parents' behalf for school-authorized educational use (16 C.F.R. § 312.5). The legal duty falls on the operator — the ed-tech vendor — which must give the school notice of its data practices plus the ability to review, delete, and stop further collection of a child's data. FTC guidance urges schools to make those notices available to parents: a best practice, not a guaranteed right.
The Children's Internet Protection Act requires schools taking E-Rate funding to filter inappropriate content and monitor minors' internet use on school networks.
Many states restrict using student data for non-educational commercial purposes and grant deletion rights. D.C.'s Protecting Students Digital Privacy Act of 2016 (D.C. Code § 38-831.01) is the local example.
Districts publish an AUP that names which laws they're complying with. DCPS's SY25-26 policy cites CIPA, COPPA, and the D.C. privacy act. Under COPPA's school-consent model the vendor owes the district notice plus review-and-deletion tools — solid ground to ask the district to share that information and exercise it for your child.
Schools sometimes assign tools below the vendor’s own minimum age. If that’s happening, it’s a fair question to raise.
YouTube's Terms require users to be at least 13 (or enabled by a parent/guardian). In 2019 the FTC settled COPPA charges against Google/YouTube for $170M over collecting children's data.
OpenAI states ChatGPT is not meant for under-13s and requires parental consent for ages 13–18. For under-13 educational use, OpenAI says an adult must conduct the actual interaction.
Microsoft sets the minimum age for Copilot at 13; the Microsoft 365 Copilot Chat FAQ states students under 13 aren't eligible.
FAQ
Straight answers, each grounded in the studies above.
Make it personal.
Take the five-minute Screenwise family survey and every guide, recommendation, and page on the site gets tuned to your kid’s age, what they use, and where your family stands.
Take the Family Survey