The panic about short-form video has an odd shape. An actor tells a trade magazine he is enraged at the hours he has handed to an Instagram feed, one of YouTube’s co-founders says he is wary of how much short-form video his own children watch, and a lawsuit against TikTok includes internal research the company would clearly rather nobody read. And yet the scientific record — the actual peer-reviewed, well-powered, replicated body of work that would tell us what these apps are doing to human cognition — is thin enough to fit on a single shelf. That gap between cultural certainty and empirical uncertainty is itself the story.
Plenty of people are convinced the app has done something to their attention. They cannot prove it. Neither, at the moment, can science.
What exists instead is a small pile of preliminary work, a much larger pile of anxious commentary, and a handful of researchers trying to keep the two apart. The Atlantic’s September 2026 survey of the field — which supplied the expert interviews and much of the reporting summarised below — lays this out plainly: the concern is reasonable, the evidence is limited, and the incentives to keep it that way are enormous.
The public voices are worth separating, because they are not the same kind of witness. Andrew Garfield, promoting a film in which a London couple move their tech-dependent children to a barn without electricity, told Variety he had given hours to nothing and that people have less control over the impulse than they should. That is a consumer’s complaint. Steve Chen, a co-founder of YouTube — whose Shorts feed recycles a great deal of TikTok — has said he is wary of his children watching too much short-form video because of what it does to attention. That one comes from someone who helped build the machine, which is a different and more interesting data point.
The most cited piece of evidence against TikTok comes from the company itself. Internal research surfaced by a sloppily redacted 2024 lawsuit filing stated that compulsive use of the app correlates with a range of negative mental-health effects, naming “loss of analytical skills, memory formation, contextual thinking” along with conversational depth, empathy and heightened anxiety. That is TikTok describing TikTok. It is also a correlation, in a document written for internal strategy rather than scientific publication, and it has not been independently replicated. A TikTok spokesperson said at the time that the complaint cherry-picked misleading quotes from outdated and out-of-context documents.
The academic picture is messier, and much of what circulates has been badly misreported. The study that went viral in August 2026 was run at China’s Zhejiang University: 56 young adults clicked through a series of short videos — not on a social platform — while lying in an MRI scanner. Two brain regions involved in executing cognitively demanding tasks were more active when participants swiped away before a video’s midpoint than when they watched to the end. That is a finding about the act of choosing to skip. It is not a measurement of intelligence, and it says nothing about depression, which did not stop one outlet running a headline announcing that Reels and TikTok make users dumb and depressed. Neuroscientists who reviewed the paper for The Atlantic pointed out that it had no control condition, treated finishing a video as evidence of enjoying it, and never established the baseline against which the supposedly video-addled brain activity was being compared.
This is the pattern worth noticing. A small study appears. A dramatic headline follows. The headline enters the cultural bloodstream. The original caveats do not.
Some emerging research does point somewhere. Caty Enders, a doctoral student at the University of New Mexico, has found that students who used TikTok at high volume had more trouble with working memory — the kind that holds information in place while you are using it — and with episodic memory, the store of specific personal events. Researchers at the University of Bristol concluded in 2025 that the constant context-switching built into short-form video may degrade prospective memory: the ability to act on something you had previously intended to do, like sending the email you meant to send.
A 2024 experiment at Sigmund Freud University in Vienna put 29 participants through a visual oddball task and found that those reporting more than four hours of short-video social media a day showed a significantly reduced P300 brainwave response, which the authors read as lower attentional capacity. Four hours is a lot. Most users do not spend four hours. The sample was 29 people. What the result means for someone spending 45 minutes on Reels before bed is genuinely unclear.
Scale is the other problem. Keith Robert Head, a Capella University researcher who recently published a review article on short-form video and attention, could point The Atlantic to only four good studies on the topic conducted in the United States. Most of the existing work comes from China, and Enders cautions against porting those findings across: American platforms operate under considerably less government control and can personalise feeds using far more user data. Findings do not transfer cleanly.
The reason the shelf is so thin is not mysterious. Rigorous cognitive research is slow and expensive, and academics are wary of spending years on apps that may be irrelevant by publication. The organisations best positioned to fund it — the platforms themselves — have no commercial reason to produce work that could expose them to lawsuits, regulation, or user flight, and have largely kept their internal research private. Independent researchers face access problems, since the algorithms and usage data sit behind corporate walls. Federal funding for potentially addictive behaviours has not been a priority. The result is a public policy vacuum being filled by intuition, celebrity confession, and the occasional viral misreading of a 56-person MRI study.
Teachers describe the same difficulty from the other end. Many are certain their students’ attention has changed over the past decade, and equally clear that they cannot separate the app effect from the pandemic effect, the phone-in-pocket effect, the sleep-deprivation effect, or the plain fact of their own accumulating impatience with teenage restlessness. The confound is the point. Human cognition sits inside a stack of variables, and pulling one out cleanly is hard.
Luiz Pessoa, a neuroscientist at the University of Maryland who reviewed the viral study, thinks an attention epidemic is plainly under way even though scientists understand it poorly. He describes watching Formula 1 highlights with his teenage daughter while she is also looking at her phone, and notes that the degree of simultaneous task-juggling would have been unrecognisable 20 or 30 years ago. That is true, and it is measurable, and it is also not the same claim as suggesting that short-form video is rewiring the brain in uniquely harmful ways. The brain is always being rewired. That is what brains do. The question is whether the rewiring caused by 133 minutes of TikTok is meaningfully worse than the rewiring caused by 133 minutes of anything else that fragments attention — cable news, group chats, open-plan offices, a toddler.
That number is not arbitrary. Participants in Enders’s study on TikTok use and memory spent an average of 133 minutes a day on the app. “If you do anything every day for 133 minutes, it will change you,” she told The Atlantic. It is a reframing more than a finding. The variable might not be short-form video specifically. It might be the sheer volume of anything replacing what those minutes used to hold.
Some of what gets blamed on the medium may simply be time displacement. Two hours on Reels is two hours not spent reading, not spent sleeping, not spent bored enough for a thought to finish itself. The harm, if there is harm, might live in the absence rather than the presence. That distinction matters for policy. Banning an app addresses one theory. Restoring the conditions for sustained attention addresses another.
There is a version of this article that would tell readers exactly what to do. That version would be dishonest. The honest position is that the concern is credible, the mechanism is plausible, the internal industry documents are damning in their own quiet way, and the peer-reviewed record does not yet support the strongest claims being made in public. Reasonable people can act on the plausibility and the internal documents without waiting for a body of research the industry has every reason to delay.
What the uncertainty protects is the platforms. What it costs is everyone else — parents making decisions about their kids, teachers watching classrooms shift, adults trying to work out whether the thing they suspect is happening to them is actually happening. Governments are in the same position, and several countries have already imposed blanket social-media bans on minors on the strength of data nobody would call sufficient. The gap between what users feel and what science has confirmed is not neutral ground. It is the same territory where a lot of modern anxiety about attention and cognition lives, and it is territory the companies have learned to occupy comfortably.
The interesting behavioural question is not whether short-form video is bad for you. It is why so many of the people who build and use these products already act as if the answer is yes. One of YouTube’s co-founders is careful about what his children watch. Actors talk publicly about hours they cannot account for. Ordinary users delete the app and reinstall it and delete it again. That collective behaviour, from insiders and users alike, is a data point of its own, even if it does not appear in any journal.
Science will eventually catch up. The slower forms of attention that used to structure ordinary life — long conversations, unhurried reading, boredom that resolved into thought — will either return with intention or continue to thin out. That is a choice being made every day, at the individual level, without waiting for the study that would settle it.
The apps are betting most people will not make it. So far, they are winning.