A six-year-old comes off the tablet after twenty minutes and can tell you that the dragon needed three coins and that the purple one is her favourite. She cannot tell you one word she read. The same child, same twenty minutes, on a different app, came off it saying "ch says /ch/, like chair" — because something on the screen asked her to say it out loud, waited, and asked again when she got it wrong.
That gap is not about screen time. It is about what the screen asked her to do.
Active and passive is a question about the child's mouth and hands
The useful distinction is not "educational content" versus "cartoons". It is whether the child produces anything that could be wrong. Watching is passive. Tapping to advance is passive, even when the thing you tap is a letter. Choosing between two plausible options, saying a sound aloud, typing a number, dragging a fraction to the right spot on a number line, explaining why — those are productions, and productions are where the learning sits.
You can measure this without any special knowledge. Sit through one session with a stopwatch mentality and count the moments your child could have got something wrong. In ten minutes, a decent phonics or arithmetic app will create somewhere between fifteen and forty of those moments. An animated story with a three-question quiz at the end creates three. Both may be perfectly pleasant. Only one of them is practice.
There is a second test that takes ninety seconds. Open the app yourself and tap at random — corners, wrong answers, whatever. If you can finish a lesson with nonsense input and still get the confetti and the "great job!", the app is not checking anything. A surprising number of highly rated titles fail this.
What the age guidance actually says
Paediatric bodies broadly agree on the youngest ages and deliberately loosen up after that. The shape looks like this:
| Age | General thrust of mainstream guidance |
|---|---|
| Under 18 months | Avoid screen media other than video calls with family. |
| 18–24 months | If you introduce it, choose high-quality content and watch alongside, naming what is on screen. |
| 2–5 years | Around an hour a day of high-quality content, less is better, ideally shared with an adult. |
| 6 and older | Most guidance drops the single number and asks you to protect sleep, daily physical activity, meals and offline play first. |
Two details get lost in the headlines. First, school-assigned work on a device is usually excluded from these figures — the numbers refer to recreational screen use. If your child's school issues a tablet and sets homework on it, you are not blowing the guidance by complying with the teacher. Second, national advice genuinely differs. Some health authorities are stricter than others for under-threes, while the education ministry in the same country may expect device-based homework from the first years of primary school. Parents receive both messages at once, and neither side is being unreasonable; they are answering different questions.
One finding is worth knowing because it explains the under-two advice. Very young children — roughly up to two and a half or three — learn markedly less from a screen than from a person doing the identical thing in front of them. The effect shrinks when the screen responds to the child specifically: a video call with a grandparent who says the child's name and waits for an answer behaves much more like a person than a recorded video does. Responsiveness, not resolution, is what closes the gap.
Co-viewing is talking, not supervising
Sitting nearby with your own phone is not co-viewing. The thing that changes outcomes is the conversation wrapped around the session, and it does not have to be long. Two minutes before, two minutes after, works.
- Before: "What are you doing today?" Making the child state the goal out loud is a small act of attention.
- During, once: pause and ask a prediction question — "what happens if she picks the eight?" Predictions force commitment, and a wrong prediction is more instructive than a right one.
- After: "Teach me the hard one." A child who can explain a step in one sentence has it. A child who says "I dunno, I just pressed it" has not.
For a four-year-old, this looks like naming what appears on screen and then pointing to the same thing in the room. For a nine-year-old it can be entirely hands-off — you are in earshot with the door open and you ask two questions at the end.
Sitting in also catches the app's mistakes, and they are common. Mispronounced phonemes, especially the schwa dragged onto consonants so that "b" becomes "buh". Content that assumes a US context: cups and ounces, state capitals, a spelling of "colour" your child's teacher will mark wrong, dates written month-first, coins your child has never held. None of this makes an app useless, but you want to know it is there so you can say "we do it differently here" rather than letting your child absorb it silently.
Telling a teaching app from an entertainment app in a school costume
Marketing copy is a poor guide. Behaviour is a good one. The most revealing moment is a wrong answer.
| Signal | School costume | Actually teaching |
|---|---|---|
| On a wrong answer | Sad noise, shows the answer, moves on | A hint that narrows the problem, a second attempt, and the item returns later |
| Difficulty | Fixed content; levels are cosmetic | Adjusts so the child sits at roughly 70–85% correct |
| Old material | Never comes back | Reappears days later, not just within the same lesson |
| End of session | There isn't one | A natural stopping point: "you're done for today" |
| Rewards | Coins, chests, characters to collect | Plain feedback and a clear next step |
| Parent view | Minutes and streaks | What was practised and what is still shaky |
Ask one more thing: does it teach a method, or only test? Plenty of maths apps are timed quizzes with a mascot. Quizzing has a place once a skill exists, but a child who cannot yet subtract across a zero will not learn it by being asked faster.
Design aimed at children rather than for them
Some mechanics exist to grow session length, and they work best on the people least equipped to resist them.
- Loss-framed streaks. "Don't lose your 43 days" lands very differently on a seven-year-old than on an adult, particularly when the app sells streak repairs. Decide in advance that streaks break during holidays and illness, and say so out loud when it happens.
- Randomised rewards. Chests, eggs, spinning wheels, mystery pets. Unpredictable payouts are the same mechanic that makes slot machines sticky. In a learning app the reward should be predictable and boring.
- Autoplay and countdowns. Removing the decision point removes the child's chance to stop. Turn autoplay off, then check again after the next update — these settings quietly reset.
- Notifications to the child's device. A character who "misses" your child is a retention tactic. Switch them off at the operating-system level, not in the app.
- Flimsy parent gates. "Tap three times" or a single-digit sum is not a gate for a numerate eight-year-old. A real one uses a PIN or a multi-digit calculation.
- Ads and "watch a video to continue". Worth checking what the ads are: children's apps sometimes serve trailers for games built on the exact mechanics you are trying to avoid.
- Wilting gardens and sad pets. Anything that punishes a day off is designed to make a break feel like a failure.
A fifteen-minute audit before you hand anything over covers most of this: install it yourself, play a lesson as a beginner, deliberately get things wrong twice, open every settings screen, and read the parent view. Kid Genius World and everything else your child uses should survive that.
The test that matters is off the screen
After two or three weeks, check whether the learning travels. Ask your child to read the sign at the bus stop, work out whether you have enough coins, name the continent the plane is flying over, tell you what a lever does. If the skill only exists inside the app — if it needs the app's exact question format, its voice, its layout — it is app skill, not a reading or maths skill, and that is worth knowing before another month goes by. Real learning shows up somewhere nobody was scoring it.