The Companies That Built Social Media Are the Only Ones Who Know the Harms It Can Do
Two academic studies suggest social media barely harms teenagers. The internal documents from Meta's own researchers tell a different story — and explain why the gap exists.
Last week, Pew Research published a survey of 1,458 American teenagers — the most current data we have on how kids think about the platforms they use. The headline finding: most teens say TikTok, Instagram, and Snapchat don’t hurt their mental health. Only 9 to 11 percent of users said these apps had harmed them.
That landed in the same conversation as a Stanford study from last year that is still circulating widely. I saw someone announce last week he was quitting social media for a year on the basis of it. That study — the largest randomized experiment ever conducted on social media deactivation, involving 35,000 participants and 27 co-authors — found that deactivating Instagram produced effects so small that the researchers’ own language described the anxiety and depression outcomes as statistically indistinguishable from zero.
Two data points. Both being used, especially by the industry, to suggest social media is basically fine for kids.
Here is the problem with both of them.
The Stanford study was conducted during the six weeks before the 2020 presidential election — a period unusual enough that its authors urged caution about generalizing from it. Eight of the 27 co-authors work for Meta. Less than one percent of people invited to participate actually completed the study. And the effects that did appear — 0.041 standard deviations for Instagram — fall roughly ten times short of the 0.41 standard deviations that researchers use as the threshold for clinical significance, meaning effects people would actually notice in their own lives.
The Pew report is more nuanced than its headline suggests. Yes, 9 to 11 percent of teens say the platforms hurt their mental health — though at the scale of billions of users, that is not a small number. More teens reported negative effects on sleep and productivity than on mental health, with TikTok leading on both. And there’s an important gap between how teenagers describe the platforms’ effect on themselves versus their peers: a separate Pew survey drawing on fall 2024 data found that 48 percent of teens believe social media is mostly negative for people their age — up from 32 percent in 2022. Teenagers may be better at seeing what the platforms do to the people around them than to themselves. Parents are more pessimistic still: 24 percent say social media hurt their teen’s mental health versus 8 percent who say it helped.
None of this, I would argue, settles the question.
The strongest recent academic study on this question is the Karolinska Institutet study, published in Pediatrics Open Science in December 2025, with no conflicts of interest to declare. Researchers followed 8,324 children from age 9 or 10 for four years. They found that social media use — specifically, not television, not video games — was associated with gradual increases in inattentiveness. The association held after controlling for socioeconomic background and for genetic predisposition toward ADHD. And crucially, children with higher inattention did not subsequently use social media more, which addresses the most obvious alternative explanation: the causal arrow runs from use to symptoms.
Over the four years of the study, average daily social media use rose from 30 minutes at age 9 to two and a half hours at age 13. Most of those children were on platforms they were legally too young to use.
This is the science that academics can produce from the outside. Recruiting kids, obtaining parental consent, relying on self-reported screen time, working with samples in the low thousands. At its best — as in the Karolinska study — it’s careful, directional work. What it cannot do is establish the causal link at scale that would settle the debate.
A researcher I spoke with some years ago described leaving a university position for a private tech company. One of the reasons he went, he said, was institutional review boards. At a university, you spend years navigating permissions, checks, consent processes — and at the end of it, you might have a few hundred participants. At a company like Meta, he said, you could study the behavior of millions of people each day. Run experiments continuously. At scales he couldn’t imagine in an academic setting.
Meta has exactly that. Behavioral log data on tens of millions of children — what the algorithm delivers, how long each user watches, what behavior follows. They can infer emotional states from those patterns; they already use this capacity to target advertising.
Researchers at NYU Stern’s Tech and Society Lab have now catalogued 35 internal Meta studies drawn from whistleblowers and litigation discovery, launched in January 2026 and updated last month. What those studies found is different in kind from what external researchers have been able to measure.
Thirteen percent of 13-to-15-year-olds reported experiencing unwanted sexual advances on Instagram every week. Eight percent were exposed to suicide-related content weekly. One internal study concluded that Instagram “makes body image issues worse for one in three teen girls” — first reported by the Wall Street Journal in 2021, citing documents provided by whistleblower Frances Haugen. In an internal study on what Meta called “severe problematic use” of Facebook, Mark Zuckerberg himself wrote in correspondence that “3 percent of billions of people is a lot of people.”
At 3 percent of three billion monthly active Facebook users, that is 90 million people Zuckerberg identified internally as struggling with severe problematic use.
From the trial documents introduced in the Los Angeles case — where a jury in March found Meta and Google liable and awarded $6 million in damages — one internal Meta memo read: “If we wanna win big with teens, we must bring them in as tweens.” Separate internal data showed that 11-year-olds were four times as likely to return to Instagram as to competing apps, despite the platform’s stated minimum age of 13.
In Meta’s legal defense, the company has repeatedly cited the external academic literature — the same literature whose limitations are a direct consequence of Meta’s data being unavailable to external researchers. On January 31, 2024, testifying under oath before the Senate Judiciary Committee, Mark Zuckerberg said: “The existing body of scientific work has not shown a causal link between using social media and young people having worse mental health outcomes.” Technically accurate. The external literature cannot establish that link.
The American Psychological Association’s chief science officer, whose work Zuckerberg cited, had this to say, however:
To support his claim, Meta later referred reporters to the American Psychological Association’s 2023 Health Advisory on Social Media Use in Adolescence (PDF, 431KB)opens in new window as evidence. We agree with Zuckerberg when he said that it’s “important to look at the science”—as psychologists, we have actually done so. And as APA’s chief science officer and a longtime researcher into the impact of social media on kids, I can say that Zuckerberg cherry-picked data to suit his purposes.
The APA advisory, which was based on the best available science, showed that social media is related to psychological harm through online discrimination, prejudice, hate and cyberbullying. Research also has found that young people face serious risks when they are exposed to content about self-harm, harm to others or eating disorders. But what is key to know is that the advisory also outlined the science behind why certain features, functions and content on social media can be harmful to young people, whose brains have not yet fully developed.
What comes next may sharpen the picture. Australia has banned social media for users under 16. Sweden has moved in a similar direction. Singapore is considering it. Those bans are imperfect policy instruments — but they will, for the first time, give independent researchers something close to a control group. Academics will soon be able to compare, say, Australian teenagers to teenagers elsewhere. The science will get better.
In the meantime, the platforms have built a system in which they are the only ones who can fully study it.
FURTHER READING
Teens’ Experiences on TikTok, Instagram and Snapchat — Pew Research Center, April 15, 2026; the news hook, with platform-specific mental health, sleep, and productivity findings
The Effect of Deactivating Facebook and Instagram on Users’ Emotional State — Stanford SIEPR/NBER working paper, April 2025; the viral study whose effect sizes are examined in detail
Digital Media, Genetics and Risk for ADHD Symptoms in Children — A Longitudinal Study — Karolinska Institutet / Pediatrics Open Science, December 2025; the strongest recent academic study, 8,324 kids tracked for four years, no conflicts of interest
Meta’s Internal Research — NYU Stern Tech and Society Lab, launched January 2026, updated March 2026; catalogue of 35 internal Meta studies from whistleblowers and litigation discovery
Teens, Social Media and Mental Health — Pew Research Center, April 2025; the prior-year survey showing 48% of teens say social media is mostly negative for people their age
Facebook Knows Instagram Is Toxic for Teen Girls, Company Documents Show — Wall Street Journal, September 2021; original reporting on the Frances Haugen documents, source of the “one in three teen girls” internal finding
Jury Finds Meta and Google Liable for Role in Young Woman’s Mental Health Issues — NPR, March 25, 2026; verdict and trial documents, including the “bring them in as tweens” internal memo