Meta and TikTok teen safety posts: 352 announcements, almost no proof they work

When YouTube, TikTok, Meta and Snapchat announce new protections for teenagers, they rarely say whether the feature is live, where it applies, or whether it works. That is the finding of a study of 352 press releases and safety blog posts the four companies published from 2019 to 2024, which researchers will present at the ACM CSCW conference in Salt Lake City on October 14.

The team, led by Pamela Wisniewski of the International Computer Science Institute, sorted every claim by what risk it addressed and how it was described. The posts were heavy on what teens see and who contacts them, and light on how the companies collect teen data or let children onto their platforms at all.

Where the attention went

YouTube contributed the most posts, 134, followed by Meta with 99, TikTok with 74 and Snapchat with 45, according to the paper.

Content exposure made up 30.7% of the topics covered and interpersonal communication 27.3%. Data access got 12.8%, content creation 6.0% and platform access, meaning age checks and sign-up, just 3.1%. Risks from how content spreads once posted made up 7.4% of the risk mentions.

The researchers argue this framing puts the work on families and teens to manage what reaches them, while saying little about the business side of the platforms.

"Rolling out" is doing a lot of work

Many announcements left it unclear whether a feature existed for a given user. Meta described its Nudity Protection feature with words like "testing," "rolling out" and "planning," the paper found, and Teen Accounts were "rolling out globally" before a later note said they may not be available in your location.

Some protections were limited to Europe without saying so plainly. Snapchat restricted personalized ads for teens in the EU and UK, and TikTok described limits covering the European Economic Area, the UK and Switzerland, with no word on the United States.

Rules also changed without explanation. TikTok's default private accounts covered ages 13 to 15 in 2021 and all teens under 18 by 2025, and Meta moved from "16 and under" to "under 18," according to the study, which found neither company explained the shift.

Numbers with nothing to compare

When companies did claim results, the figures rarely had context. YouTube said a classifier would remove twice as many predatory comments, without a baseline, the researchers noted.

"We need to be careful about equating a polished, supportive-sounding AI-generated response with meaningful support," Wisniewski said about a second study by her team on chatbots and dementia caregivers. The safety paper makes a similar point about corporate posts: a confident announcement is not evidence that a teenager is safer.