We replaced 30 minutes of YouTube with an AI game-maker. The kids didn't notice.

For two weeks in April we ran the same experiment in three different households. The deal: thirty minutes of after-dinner YouTube becomes thirty minutes of making something instead. Same screens, same window, same parents, same volume. The kids were nine, six, and four.

We did it because the YouTube algorithm had gotten weirdly aggressive about pushing unboxing videos and "challenge" content into our six-year-old's feed. The signal-to-noise on the device our kids spent the most time on was getting worse, not better. We didn't want to police every video. We wanted to see whether a different default would absorb the same time slot without negotiation.

The rules

• Thirty minutes a day. Same window the YouTube used to be: 7:00–7:30pm, after dinner, before bath.
• BrainLab Games on the same device that used to run YouTube. Same room. Same parental presence.
• No restrictions on what kind of game the kid asks for. Adult vetoes only for actual safety or sleep-disturbance reasons (no spooky bedroom monster game at 7:25pm).
• If the kid hates it on day 1 or 2, we abandon. We weren't going to force-feed creative time.
• Logbook: every session, two lines. What they made. How they reacted.

The logbook, abridged

Day 1. The four-year-old asks for "a video on the iPad" out of pure habit. Mum redirects to BrainLab Games. He dictates a prompt about a dinosaur eating a sandwich. The dinosaur eats the sandwich. He laughs for the full thirty minutes. The six-year-old builds a maze game. The nine-year-old, the hardest to convince, types a single sceptical sentence and then proceeds to make a parallax-scrolling platformer in three iterations, complete with a boss fight he names Mr. Tomato.

Day 2. The four-year-old has invented the concept of a sequel and demands dinosaur eats sandwich 2. He gets it. He has Opinions about whether the sequel improves on the original (it does, "because now the sandwich is fighting back"). The nine-year-old makes a card-matching game; complains it's too easy; asks for it to be harder; wins; demands harder again; eventually asks for "a card matching game where the cards run away when you try to flip them." This is, I want to point out, a genuinely creative game design idea, and his first one.

Day 4. The six-year-old stops asking for new games and instead spends the whole session iterating on a single one — a chase game with a kitten — making it incrementally harder, adding a second player slot, then a high score, then "make the kitten cry when you lose." We had not seen sustained iteration on anything before, ever. Her teachers had been working on it at school for a year. Apparently she just needed the right canvas.

Day 7. Mid-experiment check-in. The kids have not mentioned YouTube once. The parents have not had to redirect anyone. The four-year-old has expanded his vocabulary by, conservatively, twenty words — most of them game-design adjacent (level, boss, combo, score). The six-year-old has started saying things like "let's tweak it" unprompted.

Day 11. First negative data. The nine-year-old wants to play a regular game (Minecraft, his usual) and is mildly annoyed at the 30-minute window being "wasted" on making instead. We let him spend the slot on Minecraft instead, with the rule that he can re-enrol whenever. He goes back to making after two days off.

Day 14. Experiment ends. We ask each kid if they want to keep the new default or revert.

All three said keep it.

What the AAP actually says

The American Academy of Pediatrics' current screen-time guidance has been quietly evolving. The headline number that gets quoted — "no more than X hours" — is now bracketed by an important distinction: passive screen time (watching, scrolling) is the thing that's harmful at higher doses. Active or creative screen time — making, building, communicating with known people — is treated much more permissively. The line in their 2025 update is essentially: if your kid would be doing this with paper, paint, or Lego if the screen wasn't there, the screen version of it isn't categorically worse.

That doesn't mean unlimited. Eye fatigue is real. The transition home from "active mode" can be hard. We still cut at thirty minutes. But the moral calculus is different from "are they watching too much YouTube" — closer to "are they spending too much time in their sketchbook." A different conversation.

Things we did not expect

The four-year-old retained. He could not yet read. He could not yet type. He still managed to drive the whole session via dictation, recognised his prompts when we read them back two days later, and recognised the games he had made out of a folder of nine.

The transition to bath got easier. The dread of "stop watching, now bath" was replaced by a much less negotiable "wrap up your game, now bath." Iterators are easier to extract from a screen than passive viewers; the cliffhanger isn't external.

The siblings started playing each other's games. The four-year-old's prompts make six-year-old games; the six-year-old plays them and asks for variants; the nine-year-old occasionally consents to playing one of his sister's "for research" and then quietly makes a better version. There's a thing in there about how having to play someone else's creative output reframes how you evaluate creative output generally. I am not sure how to measure that, but I think it's real.

Caveats

Our nine-year-old is a slightly atypical case. He liked Scratch already; he was primed to enjoy game-making. Your nine-year-old might bounce off this much harder. The four-year-old, conversely, is the case I think generalises most — when the input is "talk", almost any four-year-old will engage.

Two weeks is a short experiment. The honeymoon effect is real. We're now eight weeks in and engagement is lower than the peak (closer to twenty minutes than thirty) but the displacement of YouTube is holding.

This is also one family. Three kids. We don't have a controlled sample. We're reporting what happened. If your kid hates it on day 2, your data is your data, and you should listen to it.

The take-home

The default beats the lecture. Every parent I know who has tried to "limit YouTube" has had a worse time than the parents who have substituted something. The substitution doesn't have to be school-flavoured to count as a step up. It just has to absorb the same time slot, with the same screen, with the same parents in the room, while putting the kid into a mode of doing rather than receiving.

For us, this was the substitution. Your mileage will vary. The thirty-minute window is yours to repurpose either way.

— Louis, parent of three (one is a niece in the experiment), occasional dinner-table data nerd.