Online dangers come in more forms than most parents expect. Alongside the risks that come from contact with other people, there's a quieter and equally common path to harmful content that requires no interaction at all.
A child searching YouTube for Minecraft videos gets a thumbnail of something violent in the “Up Next” panel. A preteen types a homework topic into Google and autocomplete fills in a phrase that goes somewhere dark. A video ends, autoplay kicks in, and what plays next is nothing a parent would have chosen.
These aren't edge cases. They are the everyday mechanics of how kids stumble onto harmful content. Understanding them is the first step toward doing something about it.
The recommendation problem
When a child opens YouTube and starts watching, the platform's recommendation engine gets to work. Its job is to predict what will keep that viewer engaged, and it optimizes for watch time rather than age-appropriateness.
A 2024 peer-reviewed study published in JAMA Network Open, led by researchers at the University of Michigan, analyzed nearly 2,900 video thumbnails recommended to children searching for popular terms like Minecraft and Fortnite. More than half contained what researchers described as shocking, dramatic or outrageous messaging. Nearly a third included violence, peril or pranks. And about 29 percent featured imagery the researchers classified as creepy, bizarre or disturbing.
The lead researcher, Dr. Jenny Radesky, noted that platforms with billions of hours of content cannot perform human review of everything suggested to children and instead rely on imperfect automated systems to do the work. The result is a recommendation feed that feels child friendly on the surface but frequently isn't.
The autoplay feature compounds the problem significantly. Unless a parent has disabled it, YouTube begins playing the next suggested video automatically when one ends, meaning a child does not have to choose anything — they simply sit still while the algorithm decides what comes next.
Research simulating how the YouTube recommender system operates found that users can drift from child-appropriate content to inappropriate content through a chain of automated recommendations, with the autoplay feature specifically implicated in enabling that drift.
What autocomplete does to a simple search
Search engines try to help users by predicting what they're looking for. The autocomplete feature that fills in suggestions as someone types is powered by aggregate data about what other users have searched for. For most adults, this is a convenience. For children, it can be an unexpected hazard.
A child starting to type something innocent can be shown search completions that suggest terms they weren't looking for and may not be ready to encounter. This is especially true on platforms that serve broad, general audiences, where the aggregate search behavior skews toward adults. SafeSearch on Google and similar filters on other platforms reduce this risk substantially, but they do not eliminate it entirely. The filters depend on imperfect automated classification systems.
Researchers studying autocomplete have also documented that search suggestions can surface misinformation, stereotypes and other problematic content based on what large numbers of users have previously searched, rather than what is accurate or appropriate.
Mislabeled videos and disguised content
Some harmful content reaches children not through a flaw in the system but through deliberate manipulation of it. Creators have learned that tagging videos with popular children's character names, game titles or other child-friendly terms gets those videos in front of young viewers, regardless of what the content actually contains.
Researchers have documented videos labeled with names like popular cartoon characters that depict violence, adult behavior or dangerous activities. They slip past automated moderation by exploiting the keywords children use to search. YouTube has taken steps to address this, but the volume of content uploaded every minute makes comprehensive review impossible.
Even well-intentioned videos can mislead. A thumbnail might show something funny or interesting while the actual content goes somewhere else entirely. Young children often struggle to evaluate whether content is appropriate before clicking, and may not recognize they are watching something harmful until they are already several minutes in.
The settings that help — and their limits
The good news is that all major platforms offer tools designed to reduce exposure to inappropriate content. Understanding what each one does, and what it doesn't, helps parents use them strategically.
- Google SafeSearch filters explicit content from search results, including images, videos and websites. When a parent sets up supervision for a child's Google account through Family Link, SafeSearch is turned on and locked automatically so the child cannot change it. Parents who have not yet set up Family Link supervision can do so at families.google.com. Once supervision is active, SafeSearch is enforced across all devices where the child is signed into their Google account.
- YouTube Restricted Mode is a separate control that filters mature content from YouTube specifically. It does not lock automatically in the way SafeSearch does for under-18 accounts, so parents need to enable it manually. In the YouTube app, it can be found under the account profile icon and then Settings, then General. On a browser, it appears at the bottom of any YouTube page. Because Restricted Mode operates browser by browser and device by device, it needs to be enabled on every device and browser a child uses.
- YouTube Kids is a separate app designed for younger children, with a more tightly curated content library and the option to restrict content further by age range. Parents can also disable search within the app entirely, limiting children to only what appears in the curated feed.
- Autoplay on YouTube can be turned off in the app and browser settings. This one change meaningfully reduces the risk of a child drifting from appropriate to inappropriate content, because each new video requires an active choice rather than passive acceptance of whatever the algorithm queues up.
A few important caveats: no filter is complete. Researchers and parents have consistently found that restricted modes and safe search settings reduce exposure but do not prevent it entirely. Content gets through that shouldn't, settings reset when browsers are updated, and a child using a different browser, a private window, or a shared device may bypass filters without realizing it. Technical controls work best as one layer of protection alongside conversations and trust.
Teaching kids what to do when something feels wrong
Settings protect children more reliably when children also know how to protect themselves. The U.S. Department of Homeland Security recommends that parents create a series of steps children should follow when they encounter inappropriate content — not as a punishment protocol, but as a simple, practiced response that becomes instinctive.
That plan can be as straightforward as three steps: look away, close the window, tell a trusted adult. The goal is to give children a concrete action rather than leaving them frozen or curious.
A few things are worth teaching alongside that plan:
- The gut check. Children should know they are allowed to trust when something feels wrong, even if they can’t explain why. If a video makes them uncomfortable, if the title or thumbnail seems off, or if the content shifts in a direction they didn't expect, those feelings are reliable signals. Naming this for kids in advance makes it easier to act on.
- Coming to you without fear. One reason children don't tell parents about inappropriate content they've seen is that they worry they'll get in trouble. Research and child safety experts consistently find that children who feel safe telling trusted adults about disturbing online experiences are better protected than those who keep it to themselves. Making clear that coming to you won’t result in losing device privileges makes that conversation far more likely to happen.
- Reporting and blocking. Older kids and teens can be taught to use the report and block features built into every major platform. Reporting a video or flagging a search result contributes to content moderation and reinforces that they have some agency over what they encounter.
- Not clicking out of curiosity. A thumbnail that looks alarming, a search suggestion that seems strange, a recommendation that feels out of place — these are moments where children can learn to pause rather than click. This is a skill, not an instinct, and it develops through practice and conversation.
Starting the conversation this school year
Back-to-school season is a natural time to do a device audit with your child. Sit down together, walk through the search and video platforms they use, and check the settings side by side. Make it a collaborative exercise rather than a surprise inspection.
Beyond settings, ask your child a few direct questions: “Do you know what to do if you see something that makes you feel weird? Have you ever had something weird come up when you were just trying to look something up?” The answers can open genuine conversations about how the internet works and what to do when it doesn't behave the way they expect.
The conversations and the settings work better together than either does alone. An informed child who knows to close the window and come to you is better protected than a child with perfect parental controls and no idea what to do when something slips through.
Explore more expert-sourced resources for parents at C Spire's Parents' Resource Center. Learn about Connect & Protect plans designed for kids here.