Predatory and just-plain-adult ads between kids' videos

What this actually looks like

Your kid presses play on a kids' video. Before the video starts, an ad loads. The ad is for whoever bid highest in the milliseconds it took the page to load, and the bidder doesn't see your child. They see a device, a rough geographic area, a viewing history, maybe a household-level interest segment cobbled together from cookies and prior watch behavior. If the household has any adults using the same device or the same Wi-Fi, the profile leans adult. The ad served leans adult.

The catalogue parents send us screenshots of, in order of frequency: horror-movie trailers (gory, loud, designed to grab attention). Mobile-app weight-loss ads with before/after photography. Dating-app spots. Mid-budget gambling-adjacent app ads (sweepstakes 'casinos' that aren't legally casinos). Online-pharmacy ads for medications a kid shouldn't know exist yet. Personal-injury-lawyer ads with reenactment footage. The occasional alcohol ad that slipped through because YouTube technically restricts those, but technically isn't the same as practically.

The mid-roll is the more brutal version. Your kid is eight minutes into a perfectly fine 18-minute Number Blocks compilation. The video pauses. An ad runs. The ad is a moody dating-app spot, or a haunted-house promo, or a cryptocurrency pitch. Then the kids' video resumes. Your child has now watched the dating-app spot in full attention. They were already in the focused-on-the-screen state, they couldn't scroll away, the device wasn't theirs to control.

The auction model is the part most parents don't realize is the explanation. YouTube doesn't 'choose' to serve a horror trailer before Bluey. A real-time bidding system runs an auction for that ad slot. The horror-trailer advertiser bid more than the family-friendly advertisers, or the family-friendly advertisers bid for a different segment and weren't competing for this slot. The system optimizes for revenue per impression. 'Is the viewer a four-year-old?' is not a variable in the optimization, because the system doesn't reliably know.

YouTube does have ad policies. They restrict alcohol, gambling, adult dating, and explicit content from running on videos flagged as 'made for kids' under COPPA. But enforcement depends on the creator correctly flagging the video and on YouTube correctly classifying the channel. It also depends on the ad system correctly applying the restriction in real-time. The system fails routinely. Most parents have a stash of screenshots that prove it.

There's also a category of ad that's technically not predatory but is wrong for kids in a softer way. Vacation-package ads that imply 'you should be on a beach.' Real-estate ads about second homes. Investment apps. Streaming-service ads with content their algorithm just decided to feature, which might be a true-crime show with a graphic poster. Not illegal, not flagged by anyone, but absolutely not what you queued up for your 6-year-old.

Why it's worse than it looks

Ads in the middle of children's content land differently than ads in the middle of adult content. An adult watching a YouTube tutorial knows what an ad is, knows when to look away, knows the pitch is for them or it isn't. A 5-year-old doesn't have any of those defenses. The ad is just video. The horror-movie trailer is just a scary thing on the screen. No contextual framing about how this is a different category of thing than the Bluey before it. The fear lands directly.

The cumulative exposure is the part nobody is tracking. Across the average kids' viewing week, and we mean average, not heavy, a child watching ad-supported YouTube can sit through 40 to 80 ad impressions. Most are fine. The few that aren't are concentrated, repeated, and shown at high attention. A horror trailer played eight times in a week becomes a thing your child thinks about at bedtime. You don't connect the dots because you didn't see seven of the eight.

There's a measurement problem too. Parents can't easily audit what ads their kid saw. Watch history doesn't include ad history. The platform shows you the videos, not the inserted commercials. So when your kid suddenly has nightmares about a doll or asks about a medication name they shouldn't know, you have no log to trace it back to. The exposure is invisible by design.

How Watchly takes this off your plate

The answer here is direct: Watchly's player suppresses pre-roll and mid-roll ads. Your child opens a video and the video plays. No ad load before. No interruption at minute eight. The 18-minute video is 18 minutes of the video. The dating-app spot, the horror trailer, the sportsbook pitch, the weight-loss app, none of them get a chance to load. The surface they would load into isn't there in our player.

The technical approach is a privacy-respecting iframe-based playback. We use YouTube's embeddable player in a configuration that doesn't trigger the ad break for the view. The actual video still gets served from YouTube's CDN. This is a different mechanism than ad-blocking, we're not stripping ads from YouTube.com, we're using a player surface where the ad insertion doesn't happen. The result your kid experiences is what you want: just the video.

The other half of the answer is the library structure. Even ad-free, a child on regular YouTube can still rabbit-hole into content you didn't pick. Watchly removes the recommendation sidebar entirely, autoplay only goes between videos already in your library, never out to suggested content. So the ad problem and the recommendation problem get solved by the same architecture. A player you control, on a library you control, with no ad inventory layered on top.

We won't pretend every edge case is closed. A creator who hard-codes a sponsor read into the middle of their actual video is doing something the player can't strip. That's just part of the video file. That category gets caught by our transcript filter instead. It flags 'this video is brought to you by' patterns so you can decide before it lands in your kid's library. Two different problems, two different mechanisms, both addressed.