How much screen time is reasonable over a break?

This plan puts a device in four of ten blocks, all of them in a session with an adult present and all of them making something rather than watching something. That ratio matters more than the minute count — an hour directing a build and an hour of autoplay are not the same hour. If you want the longer version of that argument, our screen time and AI guide works through it.

My child is seven. Is this too old for them?

Shorten every block and drop Day 4's morning. A seven-year-old can absolutely dictate a game plan, notice that an answer about their favourite thing is wrong, and demo what they made. What they can't usually hold is the three-attempt comparison, which needs enough working memory to keep version one in mind while reading version three.

My child is twelve and will find this babyish.

Raise the ceiling, not the structure. Let the Day 1 build be genuinely ambitious, let Day 2's expert topic be something they will argue with you about, and make the Day 5 demo an audience of more than one. Twelve-year-olds mostly object to being managed, not to making things.

Do I need to buy anything or sign up for anything?

No. Paper, a pen, sticky notes, and whatever AI tool the adult in the house already has access to. The off-screen blocks — Day 3 entirely, plus three others — need nothing at all.

Does my child need their own AI account?

No, and we'd advise against it at this age. Every tool block on this page is written for the adult's account with the adult present, which is both the safer arrangement and the one where the learning actually happens.

What if the whole plan falls apart on day two?

Then do Day 5. Genuinely — take whatever they made or half-made, and run the three-minute demo. A child explaining an unfinished thing has still practised the skill the week was for.

We're not on a break — can we use this on random days off?

Yes, and the blocks are deliberately independent for that reason. Any single block is a complete activity; the sequence just adds the arc. For a pick-one-tonight list instead of a sequence, our AI activities for kids guide has ten standalone ones.

Teaching kids about AI

Fall break activities for kids: what to actually do with five unstructured days

Fall break lands without the scaffolding summer has — no camp, no packing list, usually no warning that you are now the activities director. This is not a list of ideas to pick from; it is a plan. Ten half-day blocks across five days, each one low prep, deliberately alternating screen and off-screen, and arranged so that by the last afternoon your child can stand in the kitchen and explain something they made and why the first version didn't work. If your break is shorter than five days, there's a section below telling you exactly which blocks to keep.

How to use this plan

Each day has a morning block and an afternoon block, and each block has one anchor — one thing that has to happen. Everything around the anchor is slack. A block takes between forty minutes and two hours depending on how far your child runs with it, and the plan is built to survive a block being skipped entirely.

Three rules make the week work. Alternate the screen — no two consecutive blocks are both on a device, which is what keeps a break from dissolving into one long scroll. Let your child pick the order within a day; ownership is most of the motivation. And do the tool blocks together, in the same chair — not supervising from the doorway, actually sitting there, because the whole point of those blocks is the conversation about what the machine got wrong.

Each block below gives you the anchor, roughly what it costs you in prep, one line you can actually say out loud to start it, and what your child should be able to tell you afterwards. That last part is the real output of the week. A break where a kid made something and can explain it is worth more than a break where they did nine things and remember none of them.

Why an hour directing a build is not the same hour as an hour of autoplay

The week at a glance

Scan this first and decide which blocks you actually have the energy for. Nothing here needs a purchase, a booking, or more than about ten minutes of adult prep.

Day	Morning block	Afternoon block	Prep	Screen?
Day 1 — Make something	Plan a tiny game on paper, then direct an AI to build it — together, on your account	Playtest it with somebody who wasn't there, collect three fixes on sticky notes	Paper, pen, sticky notes	Morning only
Day 2 — Catch it being wrong	Your child writes ten facts about the thing they know better than anyone in the house	Ask an AI to explain that same thing; your child marks it out of ten and circles what's wrong	Paper and pen	Afternoon only
Day 3 — Out of the house	A noticing walk: collect or photograph ten things, no rules given in advance	Invent the sorting rules for what they collected, then break the rules with one weird case	Shoes. A bag or a phone camera	Neither
Day 4 — Make it better on purpose	The prompt ladder: ask for the same improvement to Day 1's game three times, each more specific	Cook or build something from written instructions only — no explaining, no filling in the gaps	A recipe or a flat-pack-ish task	Morning only
Day 5 — Show somebody	Finish it: last fixes, a title screen, a name	The three-minute family demo — what I made, what went wrong, what I'd do next	A phone to record on	Morning only

Day 1 — Make something that didn't exist on Monday

1. Morning — plan on paper first, then build it together (about 90 minutes) The anchor is that the plan exists on paper before anything is typed. Three boxes: what the game is, what the player does, how you win. Keep it absurdly small — a cat dodging falling socks is a good size; a role-playing game is not. Then sit down together and let your child dictate the plan to an AI and watch a real, playable thing come back. Say: "You're the director. I'm just the typing. Tell it what you want and we'll see what it gives us." They'll be able to tell you: that a vague plan produces a vague game, and that they caused the thing on the screen. Tool rule for this block: your child is under thirteen, so this happens on your account with you in the chair, and nothing personal goes into the box — no full name, no school, no photo, no address, no birthday.
2. Afternoon — playtest it on a human, off the screen (about 45 minutes) Hand the game to somebody who wasn't in the room: a sibling, a grandparent on a call, the neighbour's kid. Your child's only job is to watch and say nothing — no hints, no explaining the controls. Then three sticky notes: one thing that worked, one thing that confused the player, one thing to change tomorrow. Stick them on the fridge; Day 4 uses them. Say: "You're not allowed to help them. Just watch where they get stuck." They'll be able to tell you: that the thing obvious to the maker is invisible to everybody else — which is the entire discipline of testing, learned in one afternoon.

Day 2 — Catch the machine being confidently wrong

1. Morning — become the expert, on paper (about 40 minutes) Pick the single topic where your child outranks every adult in the house. Their game, their sport, a book series, a breed of dog, the lore of something you have never understood. They write ten true facts about it, numbered, no help. This is deliberately off-screen and deliberately before the AI sees anything — it's the control, though you don't need to call it that. Say: "Write down ten things about it that you know for certain and most grown-ups get wrong." They'll be able to tell you: what they actually know, out loud, which most kids have never been asked to inventory.
2. Afternoon — mark the machine's homework (about 45 minutes) Now ask an AI to explain that same topic to someone their age. Print it or read it on screen, with your child holding their list of ten. They circle every mistake, every thing that's technically true but misses the point, and everything it left out. Then they give it a score out of ten and say why. Say: "You're the teacher now. Mark it. Be harsh." They'll be able to tell you: that the wrong sentences sounded exactly as confident as the right ones — which is the single most useful thing a child can know about AI, and it lands far harder when they found it themselves than when you say it. Tool rule for this block: your account, you in the chair, nothing personal typed in — and that includes not telling it their age, name, or school to "make the explanation better."

Day 3 — Out of the house, no screens at all

1. Morning — the noticing walk (about an hour, plus however long the walk is) Go somewhere ordinary — a park, a block, a supermarket aisle. Collect or photograph ten things. The important part is that you give no criteria at all: not ten leaves, not ten red things. Ten things. Your child's own instinct for what counts as "a thing worth collecting" is the data set you'll use after lunch. Say: "Ten things. I'm not telling you what kind. That's the whole game." They'll be able to tell you: almost nothing yet — this block is setup, and it's supposed to feel like a walk.
2. Afternoon — invent the rules, then break them (about an hour) Spread the ten things (or the photos) on the table and ask your child to sort them into two piles, then write down the rule they used, in one sentence. Now hand them a case designed to break it. If the rule is "things that are alive," hand them a seed, a feather, a photo of a tree on a screen. If it's "things I could put in my pocket," hand them something the size of a pocket but soaking wet. Watch them patch the rule; watch the patched rule break too. Say: "Your rule is good. Here's something mean — which pile does it go in?" They'll be able to tell you: that rules learned from a handful of examples fall apart at the edges, and that this is precisely where an AI gets things wrong too. It's the most abstract idea in the week and it arrives with no device in the room at all.

Day 4 — Make it better on purpose

1. Morning — the prompt ladder, using yesterday's fridge notes (about an hour) Take the three sticky notes from Day 1's playtest. Pick the one about the thing that confused the player. Now ask for the fix three times, deleting the result in between: first "make it better," then "make the controls easier to understand," then "when the game starts, show a line at the top that says which key moves the cat, and make it disappear after five seconds." Line the three attempts up side by side. Say: "Same problem, three ways of asking. Which ask got you the thing you actually wanted?" They'll be able to tell you: that being specific is a skill, not a personality trait — and they can see the improvement they caused in the space of one morning. Tool rule for this block: your account, you in the chair, nothing personal in the box.
2. Afternoon — the literal-instructions game, off the screen (about an hour) Cook something, or assemble something, following only what is written down — and you, the adult, play the machine: you do exactly what the words say and nothing they meant. "Put the butter on the bread" gets you the tub placed on the loaf. "Mix the eggs" gets you two eggs, shells on, stirred. Your child rewrites the instruction until it survives you. Say: "I'm the robot. I only do what's written." They'll be able to tell you: why the specific version of a request wins, in a way that transfers straight back to the morning block. There is a reason this is the afternoon and not the morning: the idea sticks better the second time it shows up wearing different clothes.

Day 5 — Show somebody

1. Morning — finish it (about an hour) Finishing is a separate skill from making and most kids have never practised it. The last fixes go in, the thing gets a name and a title screen, and then — the hard part — it gets declared done. Set a timer if you have to. Say: "Twenty more minutes of changes, then it's finished and we stop." They'll be able to tell you: what they chose not to fix, and why. That sentence is worth the whole morning. Tool rule for this block: your account, you in the chair, nothing personal in the box.
2. Afternoon — the three-minute demo (about 30 minutes, including two false starts) Whoever is around — the other parent, a grandparent on a video call, a sibling forced into it. Three minutes, three questions, in this order: What did you make? What went wrong and how did you fix it? What would you do next? Record it on a phone. Do not coach, do not finish their sentences, and do not correct the bit they got slightly wrong. They'll be able to tell you: the whole week, in their own words — and that recording is the only proof of learning that has ever mattered to us. A child who can narrate their own process has understood it. A child who can't, hasn't, no matter how good the game looks.

If your break isn't five days

Breaks are not the same length everywhere, and plenty of families are working around one or two days rather than a week. The plan degrades cleanly. Three days: keep Day 1, Day 2, and Day 5 — make it, find out the machine is fallible, explain it to someone. That's the complete arc. Two days: Day 1 and Day 5, back to back. One day: Day 1's morning and Day 5's afternoon, with lunch in between. The demo is the block to protect; it's the one that converts an activity into something your child can still describe a month later.

If you have more days than the plan, don't add a Day 6 — repeat Day 1 with a completely different idea. The second build is where you find out whether the first one was luck. It is also, reliably, the one they're proudest of.

Ten standalone activities, if you want one block rather than a week

The one rule that covers every block on this page

Children in this age range should not be holding their own account on a general-purpose AI tool, and nothing on this page asks them to. Every block that touches a tool is written the same way: it happens on the adult's account, with the adult in the chair, and no personal detail goes into the box — no full name, no school, no address, no birthday, no photograph of the child, and no "he's nine and goes to…" framing to make the answer better. If you want the answer pitched at the right level, say "explain this to a nine-year-old" rather than telling it anything about yours.

Sitting in the chair is not only about safety. Half of the value in the tool blocks is the conversation that happens when something comes back wrong — the huh, why do you think it said that? — and it cannot happen from the doorway. If a block is going to happen unsupervised, swap it for one of the off-screen blocks; the week still works.

What your child can say by the last afternoon

"I made this and here's what I'd change." — Ownership plus a critique of their own work. Most school projects produce the first half and never the second.
"It sounded sure of itself and it was wrong." — Scepticism they discovered themselves, on the one topic where they genuinely outranked the machine.
"My first rule broke when you gave me the weird one." — The edge-case instinct, learned with no device in the room, and the reason AI systems fail in the ways they do.
"Asking for it properly is the actual skill." — Specificity as a learnable behaviour, with three side-by-side versions as the proof.

Fall break questions parents actually ask

How much screen time is reasonable over a break?: This plan puts a device in four of ten blocks, all of them in a session with an adult present and all of them making something rather than watching something. That ratio matters more than the minute count — an hour directing a build and an hour of autoplay are not the same hour. If you want the longer version of that argument, our screen time and AI guide works through it.
My child is seven. Is this too old for them?: Shorten every block and drop Day 4's morning. A seven-year-old can absolutely dictate a game plan, notice that an answer about their favourite thing is wrong, and demo what they made. What they can't usually hold is the three-attempt comparison, which needs enough working memory to keep version one in mind while reading version three.
My child is twelve and will find this babyish.: Raise the ceiling, not the structure. Let the Day 1 build be genuinely ambitious, let Day 2's expert topic be something they will argue with you about, and make the Day 5 demo an audience of more than one. Twelve-year-olds mostly object to being managed, not to making things.
Do I need to buy anything or sign up for anything?: No. Paper, a pen, sticky notes, and whatever AI tool the adult in the house already has access to. The off-screen blocks — Day 3 entirely, plus three others — need nothing at all.
Does my child need their own AI account?: No, and we'd advise against it at this age. Every tool block on this page is written for the adult's account with the adult present, which is both the safer arrangement and the one where the learning actually happens.
What if the whole plan falls apart on day two?: Then do Day 5. Genuinely — take whatever they made or half-made, and run the three-minute demo. A child explaining an unfinished thing has still practised the skill the week was for.
We're not on a break — can we use this on random days off?: Yes, and the blocks are deliberately independent for that reason. Any single block is a complete activity; the sequence just adds the arc. For a pick-one-tonight list instead of a sequence, our AI activities for kids guide has ten standalone ones.

Day 1 needs something your child can direct

Luchi is a voice-first family AI for kids. When they build, instead of an open chat box, your child works through a mission: plan the thing, direct the AI to build it, test where it goes wrong, and explain what they made — which is exactly the arc this week is built around. You get an update afterwards in your child's own words.

Get on the list

Browse ten standalone activities

Part of the series: Teaching kids about AI

Related guides in this series

How to explain AI to a child — Simple analogies and sample scripts by age (5–12), plus the three things every kid should understand about AI.
AI activities for kids — 10 free, screen-smart activities to explore how AI works together — no coding, no sign-up, mostly off the screen.
AI tools for kids — A parent’s shortlist of genuinely kid-appropriate AI tools — what to look for, what to avoid, and how to tell a real kid tool from a repackaged chatbot.