What learning games for kids should you build now?

Are kids spending enough time gaming for a learning game to matter?

Yes. Gaming is taking a much bigger share of young children’s screen time, so a kids learning game is entering a behavior that is already growing.

The Common Sense Census found that children eight and under still spent about 2.5 hours a day with screens, roughly unchanged from four years earlier. What changed was the mix. Average gaming time went from 23 minutes a day to 38 minutes, a 65% increase. Among young children, gaming has been taking time from older forms of screen entertainment rather than depending on a huge increase in total screen time.

Access also starts early. The same research found that 40% of children had their own tablet by age two, while almost one in four had a personal phone by eight.

For anyone deciding what to build now, that is a much better setup than trying to create a new habit from scratch. Millions of children already expect tablets and phones to contain games.

The catch is that a learning game competes with games, YouTube and other entertainment for the child's attention. Parents may care about multiplication or reading, but the six-year-old opening the iPad still has to want to play.

Will parents actually pay for educational games now?

Yes. Parents currently pay real subscription prices for kids learning apps, but they tend to pay when the educational promise is easy to understand.

Reading.com charges $12.49 monthly or $74.94 annually for a structured program built around one outcome: teaching children aged three to eight to read. Its website says more than 2.5 million parents have used the product.

Kiddopia charges $12.99 monthly or $79.99 annually. Its parent pitch is broader, but still concrete: more than 1,000 preschool activities, no advertising, no in-app purchases, offline use and up to three child profiles.

Prodigy currently charges $58.95 a year for its entry-level Math membership. The paid offer combines the child's game rewards with something much more useful to the buyer: parents can track progress and set goals.

The American Academy of Pediatrics has also become more explicit about the difference between low-quality engagement and useful digital experiences. Its 2026 policy on children's digital environments says well-designed media can support learning, while criticizing products built around prolonged engagement, excessive rewards, advertising and manipulative design.

The sweet spot is clear: children need to want the game, while parents need to see enough educational value to justify keeping the subscription.

What kinds of kids learning games are making real money now?

The kids apps making serious money today fall into two very different groups: focused learning products with a clear parent outcome, and open-ended games children genuinely love playing.

The contrast is easiest to see in current Sensor Tower estimates. Toca Boca World sits at the game-heavy end. Reading.com sits at the outcome-heavy end. Lingokids sells a large learning-and-entertainment library. LogicLike focuses much more narrowly on thinking and puzzle skills.

Sensor Tower estimates are useful directional data rather than audited company revenue, so we should treat the exact figures accordingly. Even with that caveat, the gap between the products is large enough to be meaningful.

Product	Main reason a parent buys it	Latest worldwide iOS estimate	Latest monthly downloads
Toca Boca World	Open-ended creative play	~$7M revenue	~1M
Lingokids	Broad learning + entertainment	~$3M revenue	~500K
Reading.com	Teach a child to read	~$1M revenue	~80K
Pok Pok	Montessori-style open play	~$800K revenue	~70K
LogicLike	Logic, puzzles and thinking	~$300K revenue	~100K

Reading.com is especially interesting for a small builder. Its estimated revenue is large relative to its flow of new downloads. We cannot turn that ratio into ARPU because the revenue comes from the whole installed subscriber base, but it shows that enormous download volume is not required when parents strongly value the result.

Toca Boca proves the other route: if children genuinely want to keep playing, the ceiling can be much higher.

Is the all-in-one preschool learning app already too crowded?

Yes. A new all-in-one preschool app would currently have to beat companies with thousands of activities, huge existing audiences and increasingly fast content-production systems.

Khan Academy Kids is already free. It covers early math, literacy, language, books, creative activities and social-emotional learning without ads or a subscription.

Kiddopia has more than 1,000 activities and adds new content weekly. Lingokids now says its catalogue contains more than 4,000 games, songs and shows.

The size of those libraries was already difficult to attack. AI has made the gap more annoying.

Lingokids revealed in 2026 that its internal studio can move from an idea to a playable prototype in about a week using AI-assisted production. More than one million daily active users can then try those prototypes, and only 20–30% make it into full production.

That means a startup cannot rely on producing lots of educational content cheaply as its edge. The incumbents can now do that too, with vastly better testing distribution.

Do learning games actually teach kids anything?

Yes, but recent research gives us a much more useful answer than “gamification works”: learning gains depend heavily on how the game is designed.

A 2026 second-order meta-analysis in the British Educational Research Journal combined 20 previous meta-analyses covering 688 primary studies. Gamified mathematics learning produced a moderate positive effect on cognitive outcomes, with an effect size of 0.454. The researchers also found that lower-quality studies tended to report larger effects, which is a good reason to ignore extravagant claims that any game automatically transforms learning.

Another 2026 second-order review in Educational Research Review reached a similar conclusion. Mathematics game-based learning was generally positive, but results varied considerably depending on context and design.

The most useful finding for builders came from a separate 2026 meta-analysis of 30 primary-school mathematics studies. Game type significantly changed the results, and simulation games performed best. Games combining story and social elements also did particularly well.

That pushes us toward a very different product from the usual quiz wrapper.

Imagine a child running a tiny shop. If stock costs 4 coins per unit and a customer buys 6, multiplication directly controls the transaction. If a bridge needs 12 units of material and the child has 9, subtraction tells them what they are missing. Fractions can determine how food is divided. Measurement can determine whether an object fits.

The mathematics changes what happens inside the game.

That is a much stronger starting point than answering five equations to earn a treasure chest.

Which subjects turn naturally into real games instead of disguised homework?

Math, logic, spatial reasoning and science currently give us the cleanest game mechanics, while spoken language becomes much more interesting when voice is part of the gameplay.

Arithmetic can control money, inventory and resources. Geometry can control construction. Measurement can determine whether machines fit together. Probability can influence strategy. Logic can unlock routes through a puzzle. Physics can decide whether something the child builds actually works.

Those skills can sit inside the game world without the child constantly moving between “learning mode” and “fun mode.”

Reading has extremely strong parent demand, but the underlying mechanics are harder to hide. Children still need to decode letters, blend sounds, read words and eventually comprehend text. Great design can make those activities enjoyable, but the skill itself often remains visible.

Coding also maps naturally to games, although that market has become crowded and the meaning of “learning to code” is changing quickly with AI.

If we rank the categories around what a new product can actually turn into a compelling game today, the ordering becomes clearer.

Learning area	How naturally it becomes gameplay	Parent can see progress	Competitive pressure	Opportunity now
Math + reasoning	Very high	Very high	High	Very high
Logic + science	Very high	High	Medium	Very high
Spoken language	High with voice	Very high	High	High
Early reading	Medium	Very high	Very high	Medium-high
Financial reasoning	High	High	Medium	Medium-high
Coding	High	High	Very high	Medium
Generic preschool	Low across fragmented activities	Medium	Extremely high	Low

Is math still the best learning-game category to build in?

Yes. Math is crowded, but it currently gives us the best mix of proven demand, measurable improvement and mechanics that can support a proper game.

Prodigy's scale removes one major uncertainty. In June 2026, the company reported that more than six million students from almost 78,000 U.S. schools had answered more than 840 million math questions during its school-year competition.

That works out to roughly 140 answered questions for every participating student. Children clearly can engage repeatedly with a math-centered game at huge scale.

The newer research also makes math more interesting rather than less. The evidence across hundreds of studies shows a real positive learning effect, while the newest primary-school meta-analysis specifically found simulation games performing best.

So I would avoid copying Prodigy's question-and-reward formula and go deeper into simulation, construction and strategy.

Picture a child arriving on an island with a few characters and limited resources. The player builds houses, grows food, runs shops, travels between areas and gradually expands the settlement.

At the beginning, the mathematics could be almost invisible. Count six pieces of wood. Split eight berries between two characters. Work out whether five coins are enough to buy an object.

The same world could later handle multiplication through production, division through allocation, fractions through recipes, geometry through construction, measurement through engineering, money through trading and probability through decisions with uncertain outcomes.

The continuity is important. We do not need 300 disconnected activities when the same underlying systems can become more demanding as the child improves.

Adaptive AI could quietly change prices, quantities, distances or constraints based on what the child has mastered. Two children could play the same adventure while working at different mathematical levels.

I would also avoid making the world look like school. No classroom hub, giant “Math Level 4” button or constant interruptions announcing that the child just solved an equation.

A child should get better at the game because the child got better at thinking mathematically.

Is learning to read still a good kids-app market?

Yes, but early reading is currently a much better market than it is an easy opening for a newcomer.

Reading.com gives us unusually clear evidence of willingness to pay. The company charges almost $75 annually for 120 phonics lessons and says more than 2.5 million parents have used the product. Sensor Tower's latest accessible estimate puts its worldwide iOS revenue around $1 million for the month, despite only about 80,000 new downloads.

The problem is the quality of what families can already get.

Duolingo ABC teaches phonics, phonemic awareness, comprehension, fluency and vocabulary for free. Education Development Center research cited by Duolingo found that children using the app for nine weeks improved literacy scores by 28%.

Khan Academy Kids also gives parents a free literacy path inside a much larger early-learning product.

A new reading app therefore needs a sharper idea than “phonics but prettier.” The openings I would still investigate are parent-child cooperative reading, the awkward transition from decoding words to reading independently, or a story world where spoken and written language are tightly connected.

Is voice AI finally good enough for a kids language game?

Yes. Voice AI now makes a genuinely different kids learning game possible because a young child can interact without reading menus or typing.

Buddy.ai is the clearest proof that this behavior already works at scale. Its Google Play listing has passed 50 million installs and accumulated hundreds of thousands of reviews. The product uses speech recognition and conversational characters to help young children practice English and basic learning concepts.

That is especially useful below age eight. Young children can speak long before they can comfortably navigate text-heavy interfaces.

Voice also fits language learning unusually well because speaking is the thing we actually want the child to practice.

I would push the format much further toward a game. A child might need to ask a character for directions, describe a missing object, order food, answer a creature's question, tell a robot what to do or persuade somebody to open a gate.

The story would stop progressing when communication fails, then give the child another natural chance to try.

That creates a much cleaner learning loop than tapping the correct translation out of four boxes.

Voice language games are probably the strongest alternative to the math simulation idea right now.

Should we build an AI tutor for kids?

I would use AI heavily inside a kids learning game today, but I would avoid making “AI tutor for kids” the main product.

Generic tutoring is becoming easier for large AI companies and existing education platforms to reproduce. A cartoon character connected to a language model is already close to a feature rather than a defensible company.

AI is more useful underneath the experience. It can adapt mathematical difficulty, recognize spoken answers, generate controlled variations of a puzzle and change hints when the game sees the same mistake several times.

Lingokids currently moves from concept to prototype in around a week and kills most of those prototypes after real children try them. A small team can borrow that mindset even without Lingokids' audience.

The useful question for every prototype is simple: did children want another round?

Is coding for kids still worth entering?

Broad learn-to-code games look less attractive today because incumbents already have huge reach and AI is changing what children actually need to learn about programming.

Tynker says more than 100 million children have used its platform and that more than 150,000 schools use it. Its curriculum stretches from block programming into Python, JavaScript, web development, data science and AI.

That's hard to attack with another sequence of coding lessons.

At the same time, AI coding tools are pushing the useful skill away from memorizing syntax. A child who can explain a problem, break it into steps, design rules, test what happened and fix errors may be learning something more durable than a child who simply remembers where a semicolon goes.

There could be a good new game hidden inside that shift.

For example, a child could program creatures, automate a little factory or command robots using conditions, loops and rules. When the system behaves incorrectly, the child has to figure out why.

I would market that around building and problem-solving rather than “learn Python.”

Coding remains interesting as a mechanic. As a generic educational category, I would rank it below math, reasoning and spoken language.

Are logic and science games a better opening than school-subject apps?

Logic and science may be the most underexploited part of the kids learning market because the subject can become the game itself.

LogicLike gives us one current commercial benchmark. Sensor Tower estimates roughly 100,000 monthly worldwide iOS downloads and about $300,000 in monthly iOS revenue for the product. That is far below Toca Boca or Lingokids, but it shows that parents will pay for thinking skills without needing the promise to be “your child will pass Grade 2 math.”

Chess gives us an even better example of the gameplay loop. A recent ChessKid program with Orange County Public Schools generated 95,888 student logins and 144,443 games over four months. Children were practicing planning, pattern recognition and decision-making by playing the actual game rather than completing a separate critical-thinking worksheet.

Science can work the same way.

Build a bridge and discover whether it holds. Create an electrical circuit and see which lights switch on. Change one part of an ecosystem and watch the population respond. Design a machine and discover why it cannot lift the load.

These ideas lose some of the easy parent pitch that comes with “learn to read” or “get better at math.” In return, they give us far more room to make something children would call a game without hesitation.

For a new builder, that trade is probably worth taking.

Should we copy Toca Boca's open-ended play?

We should copy Toca Boca's respect for play, while building a much narrower product around learning systems.

Toca Boca World is currently in another league commercially. Sensor Tower's latest worldwide iOS estimate is around one million monthly downloads and $7 million in monthly revenue.

The product is also revealing because children have enormous freedom. They create characters, decorate locations, run imaginary businesses and invent stories. The experience can keep going without a sequence of quiz screens deciding whether they are allowed to have fun again.

Pok Pok takes a similar idea in a smaller Montessori-inspired format. Its open-ended digital toys let children experiment without continuously searching for one correct answer. Sensor Tower currently estimates around 70,000 monthly iOS downloads and $800,000 in revenue.

Those economics make open play hard to dismiss as merely a nice educational philosophy.

For our product, I would constrain the world more than Toca Boca does. A math simulation needs rules and consequences because those rules create the learning.

But the child should still feel ownership. Let them design the shop, choose what to build, decide where resources go, create characters and solve problems in several ways.

That could give us enough freedom to make the game replayable without needing Toca Boca's gigantic volume of locations and decorative content.

Is financial literacy worth building as a kids game?

Financial literacy has excellent game mechanics, but a standalone money-learning app currently has a structural disadvantage against products that let children use real money.

Greenlight now reports 6.5 million family members. Children and teenagers on the platform have saved more than $700 million toward their goals and invested more than $75 million with parental approval.

That creates an educational loop a normal game cannot fully reproduce. A child can learn about saving, receive real allowance money, decide how much to save and watch the balance change.

Acorns Early, which absorbed GoHenry's U.S. product, has the same general advantage: lessons sit beside actual spending and saving behavior.

I would still use financial concepts inside a broader simulation.

A shop is a fantastic math game. Children can buy stock, set prices, make change, compare offers and decide whether an upgrade is worth buying. A settlement can introduce scarcity, budgets and trade. A later level can introduce simple expected value and risk.

Those concepts make the world richer while also teaching useful financial reasoning.

I simply would not make “financial literacy lessons for kids” our entire wedge unless we also had some connection to real-world money.

What age should we target first?

Ages six to nine currently look like the strongest starting range for a new math-and-reasoning game.

Children at six can already understand goals, quantities, inventories, maps, simple rules and persistent progress. By eight or nine, we can introduce much richer strategy without completely changing the product.

Parents also remain heavily involved in app choice and payment at those ages, which helps a subscription business.

Going younger creates a different design problem. Lingokids' own product research points out how little we can depend on text with pre-literate children. Instructions need to work through visuals, audio, gestures and very simple interactions.

Going much older makes the gameplay bar rise quickly. A ten- or eleven-year-old increasingly compares our product with mainstream games rather than with other educational apps.

For the math simulation, six to nine gives us the cleanest balance. I would move slightly younger for the voice-language game and slightly older for a deeper science or strategy game.

Should we sell a kids learning game to parents or schools?

We should sell to parents first and quietly make the product capable of supporting schools later.

School distribution can become enormous. As seen above, Prodigy reported more than six million children participating across almost 78,000 schools in its latest U.S. competition cycle. ChessKid has shown how school competitions can also create recurring social play.

But selling to schools changes the company we need to build.

Teachers want class management. Schools want rostering, reporting, privacy controls, curriculum alignment and administration. District adoption can involve procurement, implementation and support.

A parent can see an Instagram ad or hear about the game from another parent, start a free trial and subscribe the same evening.

That is a much faster way to learn whether children actually like our game.

I would still structure the learning data properly from the beginning. Every challenge should map to a skill. Progress should be measurable. Multiple children should eventually be easy to manage.

If teachers start using the consumer product themselves, we can then build classroom assignments, teacher dashboards and school competitions around something already proven.

How should a kids learning game make money?

A free playable core followed by a roughly $60–80 annual family subscription is the cleanest model for the type of learning game we want to build now.

Parents are already paying in that range. Prodigy Core costs $58.95 per year. Reading.com costs $74.94. Kiddopia costs $79.99.

I would make the free version generous enough for a child to establish a real play habit. A parent should know whether the child actually likes the game before being asked for a yearly commitment.

The paid product could then unlock the larger world, additional progression, several child profiles and better parent reporting.

Advertising is particularly unattractive in this category. The FTC's updated COPPA rules strengthen parental-consent requirements around sharing children's information for targeted advertising and place tighter limits on data retention.

There is also a brand reason to stay away from it. The American Academy of Pediatrics' current guidance explicitly warns about children's products driven by excessive engagement, advertising and manipulative reward systems.

A paid family product gives us a much cleaner relationship. We work for the parent and child rather than trying to maximize the value of the child's attention to somebody else.

How much content do we really need before launching?

We need far less launch content if the first learning game is built around one replayable system rather than hundreds of separate educational activities.

Trying to match Lingokids' 4,000-plus experiences or Kiddopia's 1,000-plus activities would be a terrible use of a small team's first year.

A simulation gives us different economics.

Suppose we build a shop. At an easy level, children count items and coins. Later they make change, multiply quantities, compare unit prices and work within a budget.

A building system can similarly move from shapes and measurement into perimeter, area and scale. A food system can move from counting into division, ratios and fractions.

The same mechanics can therefore support much more learning than a collection of one-off minigames.

For an initial release, ten missions children replay voluntarily would tell us more than 200 activities they complete once.

What learning game for kids should we actually build now?

We should build a math-and-reasoning simulation adventure for ages six to nine; a voice-driven language adventure and a science-and-logic sandbox are the strongest alternatives today.

The math idea has the cleanest evidence behind it.

Children are spending considerably more time gaming than four years ago. Parents already pay roughly $60–80 a year for strong educational outcomes. Prodigy has proved that math gameplay can reach millions of children, while the latest academic evidence says math games produce real learning gains and gives simulations particularly strong results.

At the same time, Toca Boca's current economics show how much larger children's engagement can become when the play itself carries the product.

I would combine those lessons into one world.

The player could run a settlement, island, expedition or tiny civilization. Children would build, trade, cook, divide resources, measure things, manage inventories and solve increasingly complicated problems. Mathematics would control what they can accomplish.

AI would quietly adjust the difficulty and create variations around the child's current abilities. Voice could eventually allow conversations with characters. Neither feature needs to dominate the marketing.

The parent view would translate play into something understandable: what the child has mastered, what remains difficult and what changed during the past week. Minutes played and points earned are far less interesting than telling a parent that their child can now add confidently within 20, understands equal sharing and has started solving simple multiplication problems.

For the second product, I would build a voice-language adventure for ages roughly five to eight. Children would progress by talking to characters rather than tapping vocabulary cards. Buddy.ai has already proved that young children will interact with voice learning at enormous scale; there is still room to push the experience much further toward a genuine adventure.

The third product would be a science-and-logic sandbox for roughly seven to ten. Build machines, ecosystems, circuits or transport networks and learn by seeing what actually happens. I like the product almost as much as the math idea, although the parent purchase reason is less immediate.

Reading comes next. The market is clearly valuable, but Reading.com, Duolingo ABC and Khan Academy Kids make the obvious approaches crowded. Financial literacy has great mechanics, although Greenlight and other products tied to real money have an advantage. Generic coding courses sit lower because the category is mature and AI is changing the skill parents should actually want children to develop.

I would stay away from the broad preschool library, the generic AI tutor and the usual quiz-with-rewards format.

If we had to choose one product and start prototyping tomorrow, the answer is the math simulation. Build one small world, one genuinely replayable system and enough adaptive math underneath it that two children at different levels can enjoy the same game.

Then put it in front of six-year-olds.

If they ask to play again, we have something worth expanding.

Rank	Learning game to build	Age	Why it looks strong now	Biggest problem to solve
1	Math + reasoning simulation adventure	6–9	Proven demand, strong new research, excellent simulation mechanics	It has to feel like a real game beside Prodigy and mainstream games
2	Voice language adventure	5–8	Voice finally fits young children well and language has a clear parent payoff	Existing AI tutors already have scale
3	Science + logic sandbox	7–10	Learning can happen directly through experimentation	Parent value is harder to explain in one sentence
4	Cooperative reading adventure	4–7	Parents pay strongly for reading outcomes	Excellent free and paid incumbents already exist
5	Economic simulation	8–12	Money creates strong strategy and math mechanics	Real-money family apps have a structural advantage