Choosing an English App for Your Child: 3 Types, 4 Tests

Short answer: Children’s English apps come in three types — ① media-based learning (videos, songs, games), ② online lessons with a human teacher, and ③ AI conversation — and they build different skills: input, real human practice, and daily mouth-time respectively. A price-column comparison can’t choose between them because they’re not doing the same job. Apply four tests instead: how many minutes the child’s mouth actually moves, how many days a week it realistically happens, whether the child can run it alone, and whether you can see what’s going on. And honestly? For most families the best answer isn’t one app — it’s an inexpensive combination.

In our home-study guide we made the case that what school English lacks is output — speaking time per child. This is the sequel: so which app do we actually install?

Sort before you compare

	① Media-based	② Online lessons	③ AI conversation
The child…	watches, listens, taps	talks with a teacher on screen	talks with an AI by voice
Builds	vocabulary, ear, familiarity	courage and experience with people	daily speaking volume
Speaking time	near zero	high (if one-to-one)	every session
Realistic frequency	daily	1–3×/week	daily
Cost range	free–¥2,000/mo	¥3,000–7,000/mo	¥600–1,000/mo
Runs without a parent?	yes	needs booking + supervision	yes

① Media-based — the best front door, and a door that never asks you to speak

Songs, phonics videos, word games: nothing introduces English as fun better, and the price of entry is near zero. The limit is structural: the screen never waits for your child’s reply. It’s the same passive-media mechanics as any video — years of it lands at “understands some, says nothing.”

② Online lessons — the real thing, rationed by the calendar

A real teacher who asks again, praises, and visibly understands — nothing else on this list substitutes for practice with an actual person. Two practical walls: frequency (booked lessons run 1–3×/week, and daily-short beats weekly-long for language), and logistics — for younger children, booking, connecting and supervising is a recurring parent job.

③ AI conversation — the daily reps, not the match

KidTalk lives here, so read this row knowing that.

The type’s virtue is frequency without friction: no booking, no waiting audience, no embarrassment, a few spoken minutes every day. Since language grows on turns, a structure where speaking happens every single session earns its place. KidTalk specifically switches language per utterance — a child can try just the words they know in English and fall back mid-sentence, which is the lowest possible on-ramp.

What it isn’t: a substitute for talking to people. The pauses, the re-asks, the face that lights up when you’re understood — that’s type ②’s job, and daily life’s. There’s no curriculum and no test prep either.

The four tests

Before the price column, ask:

1. Minutes of mouth movement. Not “exposure time” — time the child produces English sounds. Media-based scores near zero; group lessons divide by headcount.
2. Days per week, realistically. A so-so tool used daily beats an ideal tool used weekly. Picture your actual week.
3. Can the child run it alone? Anything needing a parent every session stops during your busy weeks — and restarting a stopped habit is the hardest part.
4. Can you see what’s happening? If you can’t tell what they talked about or what’s improving, you can’t decide to continue or to quit.

The honest conclusion: combine

No single app covers all three jobs. The combination that works in practice:

• Foundation: school + media-based input (free–cheap)
• Daily: AI conversation, a few minutes of reps
• Weekly: an online lesson with a human, if the budget allows

All three together stay under ¥10,000/month — and the cheapest layer, the daily one, does the most work by frequency. On a tight budget, keep the highest-frequency layer and cut from the top.

Frequently asked questions

From what age? Media-based: often 2–3. AI conversation: once a child can ask a question by voice, around 3–4. Online lessons: once they can sit with an on-screen teacher for thirty seconds — enormously variable.

Why no named-app comparison? Two reasons. Named comparisons go stale in months as prices and features shift. And we’re a participant in type ③ — our ratings of competitors would deserve your skepticism. The type-sorting and the four tests are the durable tools; with them you can evaluate any app on the shelf yourself.

My child quits everything. Which type sticks? Session length predicts survival better than type. Under five minutes with a clear ending sticks; a 30-minute lesson is a wall for younger kids regardless of quality.

Will an app fix pronunciation? A clear daily model helps, but pronunciation grows from many inputs — and at this age the real goal is willingness to open the mouth, which correction kills faster than anything.

In summary

• Three types building three skills — never compare them in one column
• Media = input · human lessons = the real thing · AI = daily reps
• Four tests: mouth-minutes, weekly frequency, runs-alone, parent visibility
• The best answer is usually a combination — and the cheapest layer works hardest

When in doubt, return to the one question that settles it: “Will this make my child’s mouth move?”