Mental health warnings made teens less interested in using social media than standard break reminders in a new randomized trial.
Almost every American teenager is on social media. Four states have adopted policies requiring social media platforms to show health warnings or notices to young users. Until now, nobody had tested that wording on the people it targets.
In a randomized trial at Stanford Medicine, teenagers and young adults rated every warning they saw as more discouraging than a neutral message about seatbelts. The messages naming mental health harms rated highest of all.
The sharpest split wasn’t between one warning topic and another. It was between a health warning and the take-a-break nudge that platforms already show on their own. What the trial measured, though, is what a young person predicted about themselves, not what they went on to do.
Anna Grummon, an assistant professor of pediatrics at Stanford Medicine, led the work with senior author Thomas Robinson.
“We haven’t known what warnings for social media should say, or if they’re promising for changing behavior,” Grummon said. “Sometimes, subtle changes to warnings really matter.”
States still have to write the warnings
In 2024 the US surgeon general called for social media platforms to display mandatory warnings, the way cigarettes and alcohol carry them. More than 40 state attorneys general backed the idea. California, Colorado, Minnesota, and New York have since adopted policies requiring warnings, and Texas has proposed a bill of its own.
Only California put the words in the statute. Its warning law spells out one message and takes effect on January 1, 2027. The other three hand the wording to state officials or expert panels, so somebody still has to write it.
Warnings on cigarettes, alcohol and unhealthy food have been studied for decades. A warning on a feed had never been shown to a teenager and scored, and that gap sits under the social media rules now moving through state legislatures.
Young people rated 15 warning messages
In December 2025, 1,012 young people took an online survey. Roughly half, 499 of them, were 13 to 17 years old, and 513 were 18 to 29. Each person saw all 15 messages, in random order.
Thirteen were health warnings, covering six subjects. Those were addiction, negative body image, sleep, depression, and anxiety, overall mental health harms, and the fact that social media hasn’t been proven safe for young people.
One was a screen time break message like the ones TikTok already shows. The last was a control about wearing a seatbelt, picked because it’s a health topic with nothing to do with scrolling.
Each message sat on a mock Instagram page with the background blurred, white text on black with a yellow warning triangle, filling about half the screen. Half the screen is the midpoint of the two warning sizes California’s law allows.
Volunteers rated every message from 1 to 5 on one question: how much does this discourage you from wanting to use social media? A second question asked how much it raised their awareness of the harms. Grummon’s team registered the design before collecting any data.
Mental health warnings scored highest
Every social media health warning scored higher than the seatbelt control message.
The warning about overall mental health harms finished first, at 3.04 out of 5, with depression and anxiety a hair behind at 3.00. Both gaps over the control are too big to pin on chance, and both are small in size, which the researchers say plainly.
Grummon won’t claim to know why those two won.
“If I were to speculate, I would say that these are harms that resonate with people,” Grummon said. “People may think, ‘That’s right, I can see that happening to my friends or to me,’ and these are outcomes people care about.”
Depression and anxiety are also where the evidence on heavy scrolling is strongest, particularly for teen girls.
Break reminders scored lowest
Platforms have their own answer to this question, and the team tested it on the same page as everything else. TikTok shows a screen time break message, and a version of that message went in alongside the health warnings.
It came last, at 2.74. Every health warning outscored it except the one about addiction, which landed 0.07 of a point ahead on the 5-point scale, close enough that chance can explain it.
Robinson, a professor of pediatrics at Stanford Medicine, said the two age groups moved together in a way the team hadn’t expected.
“If you’re a few years older, it makes a big difference in your experience with digital media since the landscape is changing so quickly,” Robinson said. “But our teen and young adult participants consistently said they felt these messages would reduce their social media use.”
How far the warnings beat the control didn’t vary with age, though the trial wasn’t built to test that. It matters because several state laws cover minors alone while compulsive use shows up well into the twenties.
California’s longer warning scored well
California’s message is the wordiest of the lot, a full sentence about benefits for some young users, significant mental health harms, and a product not proven safe. Young people rated it on a par with the short, blunt warnings anyway.
“We talked to the people who created the bill, and they told us the rationale behind it,” Robinson said. “It’s not the message we would have designed, but we think we understand why it performed well, and it was meant to stand up to legal scrutiny. All these bills are in litigation, being challenged by the social media industry.”
Courts have struck down warnings before for leaning on causal language judged too strong, or for naming a harm that isn’t settled science. The team wrote its messages with different degrees of causal wording on purpose, and checked none of them against that legal test.
Everything here rests on a prediction a person makes about their own future scrolling. That kind of rating tracks real message effects for tobacco and sugary drinks, and it has not been checked against behavior for a warning on a social media app.
Volunteers saw each message once, on a static post, with the feed behind it blurred out. A real warning would arrive over and over, on top of video and an algorithmic feed. Every message tested here named the surgeon general, which not every state plans to do.
The behavioral test is the trial nobody has run yet, and it’s the one the laws are written on. Grummon worked on the study with Robinson and with colleagues at the UNC Gillings School of Global Public Health and UC Davis. Their next step is a longer test of whether the warnings change screen habits at all.
Grummon said the product itself is the complication.
“With substances like tobacco and alcohol, we find, on average, that well-designed warning labels change behavior,” Grummon said. “But social media is a totally different product from cigarettes or alcohol or sugary drinks, where you just put the warning on the product and it lives there.
“Social media laws are written so that these warnings are basically pop-ups. We need to find out more about how that changes behavior.”
The full study was published in the journal JAMA Health Forum.
—–
Like what you read? Subscribe to our newsletter for engaging articles, exclusive content, and the latest updates.
Check us out on EarthSnap, a free app brought to you by Eric Ralls and Earth.com.
—–
