“Brainrot” started as a joke. It is now a word with an fMRI scan attached to it.
Researchers at Zhejiang University put 56 people in a scanner, showed them short videos, and watched two of the brain’s main self-control hubs go quiet the moment a clip actually landed. The finding is published in NeuroImage, and it is more interesting and more limited than the headlines suggest.
The short version
- Who: a team led by Tiantian Hong at Zhejiang University, published in NeuroImage
- Sample: 66 recruited, 56 usable, healthy young adults averaging about 23 years old
- Finding: the dorsal anterior cingulate cortex and dorsolateral prefrontal cortex deactivated during videos participants liked and watched to the end
- The twist: people with higher resting glutamate in the dACC showed less of that shutdown
- What it does not show: permanent damage, long-term effects, or anything about children
What they actually measured
The design here is better than the usual screen-time study, which is why it is worth reading past the headline.
Before anyone watched a single video, participants went through proton magnetic resonance spectroscopy. That is a way of estimating the resting concentration of specific neurochemicals in a specific patch of brain tissue, in this case glutamate and GABA in the dorsal anterior cingulate cortex. Glutamate is the brain’s main excitatory transmitter. GABA is the main inhibitory one.
Then they ran a short-video viewing task inside an fMRI scanner. Crucially, participants could skip. That gave the researchers a natural split in the data: clips a person chose to watch all the way through, versus clips they bailed on early. The comparison is not “phone versus no phone,” it is “a video that hooked you versus one that did not.”
The result: the brakes come off, but only for the good stuff
Two regions did something notable. The dorsal anterior cingulate cortex, which is heavily involved in monitoring conflict and flagging when something needs more control, and the dorsolateral prefrontal cortex, which does a lot of the deliberate, effortful regulation. Together they are core hubs of what neuroscientists call the cognitive control network.
Both deactivated. Not during short video watching in general, but specifically during preferred videos watched to completion. Clips that got skipped produced far less of the effect.
Schematic illustration. Region positions are indicative, not anatomically precise.
In the authors’ own words, the pattern “shifts neural processing toward emotional engagement and immediate gratification, contributing to excessive use and diminished self-control in some individuals.”
Read that carefully. The claim is not that your prefrontal cortex is being damaged. It is that while a good clip is playing, the part of you that would normally say “okay, that is enough” is turned down. Which, if you have ever looked up from your phone and lost forty minutes, will feel familiar.
The glutamate finding is the genuinely new part
Here is what separates this from the dozen other “social media and your brain” studies published this year. The pre-scan chemistry measurement predicted something.
Participants with higher resting glutamate in the dACC showed less deactivation in both regions during preferred videos. GABA showed no significant relationship. In other words, there appears to be a baseline neurochemical difference that tracks with how hard the format pulls you under.
Correlation, and the authors say so
The glutamate link is correlational. Nobody manipulated anyone’s brain chemistry. It could mean glutamate levels shape susceptibility, or that both are downstream of something else entirely, like sleep, stress or long-term viewing habits the study did not measure.
One more wrinkle the coverage mostly skipped: functional connectivity between the dACC and dlPFC actually increased during video watching, more so for preferred clips. Two regions talking to each other more while both are less active is not a simple story, and the authors explicitly warn against reading it as evidence of stronger self-control.
Four things this study does not say
| The headline version | What the paper supports |
|---|---|
| “TikTok damages your brain” | A temporary deactivation during viewing. No structural damage was measured, and none was claimed. |
| “Your attention span is permanently shorter” | Nothing here is longitudinal. There is no before-and-after over months or years. |
| “This is what it does to kids” | The sample was adults averaging about 23. No children or teenagers were scanned. |
| “Proof that short video is addictive” | Participants were healthy adults with no identified addiction or compulsive-use profile, and compulsive use was not systematically measured. |
What it does establish is a mechanism worth taking seriously: the format is not neutral. A feed that lets you discard anything that does not grab you, instantly, is a machine for finding the exact clips that produce this response. The skip button is not a convenience. It is the selection process.
Scale is what turns a lab result into a problem
Forty seconds of reduced cognitive control is nothing. The question is how many forty-second windows a person stacks in a day.
Ninety-five minutes a day, mostly delivered in clips of fifteen to sixty seconds, is somewhere north of a hundred individual hits. Reels make up around half of all time spent on Instagram. YouTube Shorts clears 200 billion views a day. The format did not just win, it reorganized every platform around itself, which is part of why YouTube is redefining what counts as a view this month.
This fits a pattern researchers keep finding
The Zhejiang result slots neatly next to other 2026 work on how recommendation systems shape behavior. A PNAS study published earlier this month found that the posts you angrily reply to on X are exactly the posts the algorithm gives you more of. Different platform, different mechanism, same underlying design: measure engagement, optimize for it, and let the consequences fall where they fall.
Policymakers have noticed, and are struggling. France passed Europe’s first under-15 social media ban and then watched its own Constitutional Council strike the law down on rights grounds. The evidence base keeps growing faster than the legal tools for acting on it.
What to do with this, practically
- Treat autoplay as the actual product. The study’s effect keyed on clips watched to completion. Anything that removes the decision to continue removes the moment where control could reassert itself.
- Timers work better than willpower. If the region that says “stop” is turned down while you are watching, the stop has to come from outside the loop.
- Watch the context, not the total. Twenty minutes on a couch and twenty minutes at 1 a.m. are the same number and very different behaviors.
- Do not over-read it. This is one study, 56 adults, one session. It is a good result. It is not a diagnosis.
What to watch next
- Replication in teenagers. Every policy argument in this space is about minors, and every good scan so far has been run on adults.
- Longitudinal work. The open question is whether repeated deactivation changes anything that persists after the phone goes down.
- Whether glutamate holds up. If baseline neurochemistry really does predict susceptibility, that is a much bigger finding than the deactivation itself.
- How platforms respond. Historically, the answer to research like this has been a screen time dashboard and a press release.
