Can you actually keep your kids safe online? A new report says don't count on it

I have two small kids. And like most parents raising children in this era, I think about screens constantly. Not just how much time they'll spend on them, but what happens the day they get their first phone, their first account, their first login. We tell ourselves there's time to figure it out.

But a new report out this week makes a pretty convincing case that even when that day comes, and even when we do everything "right," the safety nets we're told exist might not actually catch anyone.

The report, called "Broken, Buried, or Missing," comes from the Cybersafety Research Center, a joint initiative between NYU and Northeastern University, with support from the nonprofit Heat Initiative.

Researchers spent seven months, from December 2025 to June 2026, testing the actual child safety features on Instagram, Snapchat, TikTok, and YouTube. Not reading the press releases and taking them at face value. Actually creating test accounts, posing as both kids and adults, and trying to trigger every safety tool these companies claim to offer.

The topline number is the one that should stop every parent scrolling: of the 86 safety features they tested, 51 failed. That's roughly three out of every five.

How bad is it, platform by platform?

· Snapchat: 73% failure rate, Instagram: 66% failure rate, YouTube: 55% failure rate, TikTok: 50% failure rate

To be clear, a "failure" here doesn't just mean broken. Researchers found features fell into a few different buckets: some were "broken," meaning a teen could find and turn them on but they simply didn't do what they promised. Some were "buried," meaning they technically worked but were hidden so deep in settings that no realistic teenager would ever find them.

Some were both. And nine features were straight up "missing," meaning the company advertised them but researchers could never get them to activate at all, no matter what they tried.

What are these "safety features" anyway?

Before getting into what went wrong, it's worth explaining what we're actually talking about, because most parents have heard these terms without ever really knowing how they're supposed to work.

·Search restrictions: When a kid types something dangerous into the search bar, like terms related to self-harm or eating disorders, the app is supposed to block the results and redirect them to a hotline or support resource instead.

·Hidden Words: A tool that lets a user build a personal blocklist of offensive or harmful words, so if someone tries to comment or message them using those words, it gets filtered out automatically.

·Message initiation restrictions: Rules meant to stop adults from starting private conversations with teenagers they don't already follow or know.

·Time limits and "take a break" prompts: After a set amount of daily use, the app is supposed to nudge, or in some cases actually stop, a teen from continuing to scroll.

·Default private accounts: New teen accounts are automatically set so only approved followers can see their posts, rather than defaulting to public.

·Comment and conduct tools: Features meant to detect bullying, harassment, or explicit language in comments or DMs, and either block it outright or prompt the user to reconsider before posting.

· Resource redirection: When a user searches something like "suicide," the app is supposed to instantly surface crisis hotlines and support resources instead of search results.

On paper, that's a genuinely solid toolkit. The problem, according to this report, is what happens when you actually try to use them.

The findings that actually made my stomach drop

A few examples in this report go beyond "the settings menu is confusing." They get at something much scarier: kids actively searching for help, or trying to avoid harm, and the platform pointing them in the opposite direction.

On TikTok, researchers created a test account registered to a minor and had it search for content related to disordered eating and self-harm. Every platform promises that these kinds of searches get blocked and redirected to crisis resources. Instead, TikTok's search function started actively recommending related harmful search terms of its own accord, not terms the researchers went looking for, but terms the app itself suggested once it picked up on the pattern.

Instagram had a similar failure, just with a different mechanism. As researchers began typing sensitive search terms, the app's own autocomplete offered up deliberately misspelled versions, the exact kind of misspellings that online communities use to dodge content moderation. In other words, the platform's own tools taught the workaround.

Then there's messaging. Snapchat promises adults cannot find or message underage accounts. In testing, researchers using an adult account were able to directly search for, find, and message a child test account with zero restrictions, and the child could see the entire message history immediately upon accepting the request. Instagram's messaging protections held up a little better, but had a real gap too: while an adult can't initiate a conversation with a teen who doesn't follow them, if the teen messages that adult first, the adult can then respond completely without restriction. Meta told researchers this is working as intended, since the platform reads that as the teen wanting to connect.

And then there's the stuff that's almost darkly funny if it weren't so serious. Every major platform advertises a "time to take a break" prompt, essentially a pop-up that's supposed to interrupt a scrolling session once a preset daily limit is hit and nudge the kid to close the app. In practice, every one of them comes with a built-in snooze button. Instagram and YouTube let a teen dismiss it instantly. TikTok makes them wait a whopping five seconds. YouTube's version for Shorts even includes a direct link to the settings page where a kid can just turn the whole limit off. Researchers describe this as a "failure by design," meaning the feature isn't broken by accident, it was built this way.

Why does this keep happening?

The report draws a clear line between two kinds of failure. Some tools failed because of bad execution, glitches a quality assurance team should have caught. But a lot of the worst failures were failures of design. The tools were never built to actually stop the underlying behavior. They were built to look like a safeguard while quietly leaving the door open.

The category that came out worst across every single platform was what researchers call "Conduct," meaning tools meant to govern how kids treat each other, like bullying detection and comment moderation. Not one working example existed anywhere in this audit. Screen-time and compulsive-use tools weren't far behind, succeeding less than a third of the time.

Was anything actually good?

Yes, and this matters, because it proves this isn't an unsolvable problem. TikTok's "TikTok for Younger Users" mode, automatically triggered for anyone under 13, was singled out as a genuine success story. It doesn't try to filter a dangerous experience, it replaces it entirely: no open search, no messaging, no algorithmic feed, just curated, view-only content. Instagram's default-to-private setting for new teen accounts was another bright spot, protecting kids before they ever make a choice at all.

The researchers' bigger point is that the platforms that succeeded did it by removing the risk, not by trying to filter it after the fact. That's a fundamentally different design philosophy, and it's one that clearly can be built when companies choose to.

So, can we actually keep our kids safe online?

Reading through this, honestly, my answer is: not by relying on the platforms alone. Not right now. This report makes it painfully clear that the safety features we're told exist are, in large part, a kind of theater. Something to point to in a press release, not something built to withstand a curious, resourceful teenager, let alone a predator looking for a gap.

That doesn't mean we're powerless. It means the responsibility can't sit entirely with a toggle switch buried five menus deep. It means real conversations with our kids, real oversight, and treating any "safety feature" a platform advertises as a starting point, not a guarantee.

As a mom, that's not a comforting conclusion. But I'd rather know it now than believe a promise that, according to this research, three out of five times, simply isn't true.