Why Evaluating Educational Apps Takes More Than a Star Rating

Evaluating educational apps for children starts with a figure most parents never see: of the 9,000-plus apps claiming to support early reading development in major app stores, a 2025 systematic quality appraisal found only 27.5% were worth recommending, and consumer star ratings showed no consistent relationship to which apps made the cut.

Category labels like “Educational” are self-reported in both the Apple App Store and Google Play, not independently certified. The metrics that lift an app’s ranking (downloads, session length, user ratings) measure engagement. Parents who want a genuine quality signal need a different set of questions.

A Quality Problem Hidden by Star Ratings

The 2025 study, a systematic quality appraisal of 309 reading apps using the Mobile Application Rating Scale (MARS, a validated research instrument for assessing mobile software quality), found that 75% of reviewed apps scored in the poor-to-acceptable band and only 24.7% hit the threshold for “good.” Across the research on educational app quality, consumer star ratings have shown no consistent relationship to MARS quality scores.

App store infrastructure explains much of this gap. Apple’s Kids category imposes real requirements (no third-party ads or analytics, mandatory parental gates for purchases, human review of all advertising content), but developers who want to avoid those restrictions can list their app outside the Kids category while still marketing it with educational language. Google Play’s Teacher Approved badge, which appears on some app descriptions, doesn’t require an education focus at all. App listings routinely describe an app as “engaging” or “interactive” with no specification of learning outcomes or pedagogical approach.

• 9,000+ apps in major stores claim to support early reading development alone
• 75% of the 309 appraised apps scored poor-to-acceptable on the MARS quality scale
• 27.5% were recommended by expert reviewers after a full pedagogical assessment
• Consumer star ratings show no consistent correlation with expert quality scores across the research

The store’s design produces this result: it can rank apps by popularity, but has no mechanism to rank them by learning effectiveness.

The Business Model Behind Free Learning Apps

Most children’s learning apps are free to download. How they generate revenue matters before any evaluation starts.

Advertising, freemium gating that locks useful content behind a paywall, subscription upgrades, and data monetization are the most common revenue structures. An investigation by SafetyDetectives, a cybersecurity research group, analyzed 20 top-rated children’s apps and found that many harvested sensitive data for what the report described as “a lucrative surveillance economy,” despite legal protections designed to prevent it. Apps collected device identifiers, behavioral data, and in some cases location information, passing it to advertising networks through embedded third-party software development kits (SDKs).

Federal enforcement actions in 2025 illustrated how widespread the problem had become. In January, Cognosphere, the developer behind Genshin Impact, paid $20 million to settle Federal Trade Commission (FTC) allegations that included collecting personal data from children without parental consent. In September, Disney agreed to pay $10 million after allowing children’s data to be collected and shared with advertising partners via YouTube videos. That same month, Apitor Technology, which sells programmable robots for children, settled a separate case after an SDK embedded in its companion app was found collecting children’s personal information without parental notice.

A well-designed app with strong learning content can still carry a data-collection problem in its backend. The privacy policy shows developer intent; the SDKs embedded by third parties may collect beyond what that policy describes.

What Research Shows Effective Apps Do

A study published in the peer-reviewed journal Child Development in 2025 tested whether high-quality learning apps at home improve children’s literacy and numeracy outcomes. The intervention involved roughly 500 children across two German cohorts. Results were skill-specific: apps focused on letter knowledge improved letter knowledge; math apps targeting counting improved counting. The match between app content and learning target was the consistent predictor of gains.

What separated the effective apps was design: whether the app required children to produce responses, fix errors, and demonstrate understanding, or whether it let them consume content passively. The table below maps those research signals to five criteria a parent can apply without a formal research instrument.

Few apps are strong on all five. Two or three that hold up is more than any star rating ever told you.

Children’s Privacy Rules, Updated

The Children’s Online Privacy Protection Act (COPPA) Rule received its first significant revision since 2013 in April 2025, with a compliance deadline of April 22, 2026. That deadline has now passed.

Those 2025 amendments expand what counts as protected personal information to include biometric data (fingerprints, facial recognition patterns, and voiceprints) alongside device identifiers, location data, and existing categories. Companies must now obtain separate parental consent before sharing children’s data with advertisers or using it to train AI systems. The rule sets stricter data retention limits and requires privacy policies to identify specific third-party recipients by name, not broad categorical references.

The 2025 enforcement record (Cognosphere, Disney, Apitor) reflects what the FTC characterized as systemic governance failures, with advertising trackers and third-party SDKs embedded in children’s apps without adequate legal oversight. Violations now carry civil penalties of $53,088 per incident, as adjusted for inflation in early 2025. For parents, the practical implication is that apps collecting behavioral data or voice recordings from children under 13 without parental consent are more legally exposed than they were two years ago, and the updated rules give parents explicit rights to review and request deletion of their children’s data.

Parents outside the United States should check local rules. The UK’s Age Appropriate Design Code and the EU’s General Data Protection Regulation (GDPR) impose separate requirements, and enforcement varies by jurisdiction.

Where AI Features Change the Calculation

AI-powered features are appearing in children’s learning apps at an increasing rate, and they introduce an evaluation problem a standard quality checklist doesn’t fully capture. Traditional app content is fixed: a wrong phonics rule stays wrong, but it won’t generate new errors on demand. Language model-based tutoring can, and does.

AI tools used for language practice, writing feedback, or personalized pacing can genuinely improve a learning experience. They can also produce confident and plausible incorrect explanations. A child using an AI tutor for math may receive feedback that sounds authoritative and contains an error. Depending on the app’s privacy terms, the child’s typed or spoken responses may also contribute to model training data, a category of data use now requiring separate parental consent under the 2025 COPPA amendments.

Three specific questions worth adding to any evaluation of an AI-powered app: Does the app label responses generated by AI as such? Can AI features be disabled in parental settings? Does the privacy policy address what happens to the child’s prompts and interactions? Any app that markets AI learning features but doesn’t address these points in its documentation warrants a direct inquiry to the developer before a child starts using it.

A Seven-Day Test Before You Keep Any App

No evaluation happens reliably in a single download session. A structured week-long trial catches what a store listing never surfaces.

1. Before day one (parent review): Read the privacy policy for third-party data sharing and a deletion option. Identify the monetization model and check whether in-app purchases sit behind a parental control.
2. Days one and two (co-use session): Sit with the child. Watch how they navigate and ask what they’re doing. Can they explain the task in their own words?
3. Days three and four (independent sessions): Let the child use the app briefly without help. Notice whether the learning task or the reward system draws more attention.
4. Day five (offline check): Ask the child to demonstrate the skill away from the screen. Improved phonics recognition, more accurate arithmetic, or new vocabulary in conversation are the signals that matter.
5. Day six (progress review): Check the app’s dashboard critically. Accuracy rates and skill-gap data are meaningful; streak counts and badges are engagement metrics, not learning indicators.
6. Day seven (keep, limit, or delete): Keep the app if the offline test showed real gains and the routine fits. Limit it if there’s value but the reward design needs tighter boundaries. Delete if the offline test produced nothing or every session ends in conflict over stopping.

Day seven produces a decision backed by actual observation, which is more than the store’s marketing page can claim.

Frequently Asked Questions

How Can I Tell If an App Is Genuinely Educational?

Check whether the app targets one specific skill and whether the child practices that skill by solving problems, explaining answers, or correcting mistakes. After a week of regular use, ask the child to demonstrate the skill away from the screen. Improved performance offline is the clearest signal the app produced real learning.

What Data Do Educational Apps Typically Collect From Children?

Common categories include device identifiers, usage behavior, names, email addresses, and in some cases location data or voice recordings. Under COPPA, apps directed at children under 13 in the United States must obtain verifiable parental consent before collecting personal information. The 2025 amendments added biometric data (facial recognition patterns and voiceprints) to the protected list and require privacy policies to name specific third parties receiving that data.

Are Free Educational Apps Safe for Children?

Some free apps are well-designed and privacy-conscious; others rely on advertising revenue, subscription upsells, or data monetization. Apps using advertising revenue are more likely to embed SDKs that collect behavioral data as a byproduct. Before allowing a child to use any free app, check the privacy policy for mentions of third-party data sharing and confirm there’s a mechanism for parental review and data deletion.

What Should I Look for in an App’s Privacy Policy?

Look for plain-language statements about what data is collected, whether it’s shared with advertisers or third parties, how long the app retains it, and how parents can request deletion. A policy naming specific third-party recipients is more informative than one referring vaguely to “trusted partners.” The absence of a privacy policy, or one that can’t be found without extensive searching, fails the basic legal threshold set for apps marketed to children in the United States.

Do Star Ratings in App Stores Reflect Educational Quality?

Research suggests they don’t. The 2025 systematic appraisal of 309 reading apps published in the Early Childhood Education Journal found consumer star ratings had no consistent relationship to quality scores from expert reviewers using the Mobile Application Rating Scale. Apps rated highly often earned those ratings for being engaging or easy to use, neither of which predicts whether a child learns anything.

What Questions Should I Ask Before a Child Uses an AI-Powered Learning App?

Ask whether the app labels AI-generated responses as such, whether AI features can be disabled in parental settings, and what happens to the child’s inputs. Under the 2025 amendments to the federal child privacy rule, using a child’s data to train an AI system requires separate, explicit parental consent. Any app advertising AI learning features that doesn’t address data use in its privacy policy should be queried directly with the developer before use.

An app that earns a place on the device should be able to clear the bar the store listing never required.