What happens to your child’s data in a parenting app

A baby tracker holds a strange kind of record. Individually, none of it is dramatic — a feed at 4pm, a nap that ran 40 minutes, a weight in kilos. Together, over a year, it is a precise description of a household: when the adults are awake, when the house is empty, a child’s name and exact date of birth, and a health history nobody else has.

That is not a reason to avoid tracking. It is a reason to spend ten minutes checking before you commit two years of it to a company you have not heard of.

The three questions

Everything worth knowing collapses into these:

• What leaves your phone? Everything, or only what is needed to sync?
• Who else receives it? Which named third parties, and for what?
• Can you get it deleted? Permanently, yourself, without asking anyone?

An app that answers all three plainly is usually fine even if the answers are not perfect. An app that is vague about the third one is the one to skip, because deletion is the only control you have left once you have changed your mind.

Where the real answers live

Not in the marketing copy. “Bank-level encryption” and “your privacy matters to us” are decoration; neither states who receives your data. Three places actually tell you something:

The store labels are self-declared, so they are a starting point rather than an audit — but a label that contradicts the privacy policy is a genuine red flag, and the two are easy to compare side by side.

A calculation is not the same as a request to a model

This distinction matters more than any other in an AI-branded parenting app, and almost nobody spells it out.

Statistical features

A nap prediction, a growth centile, a weekly average. These are arithmetic over logs the app already holds. Nothing new leaves, no external model is involved, and the “AI” on the listing describes a formula.

Assistant features

A chatbot, a recipe generator, a story writer. These cannot work locally on a phone. What you type — and any context the app attaches to it — is sent to a language model, usually run by a third party such as Google, OpenAI or Anthropic. That is not inherently bad; it is simply a different data path, and you should know which features use it.

The practical consequence is that you control the second category by controlling what you type into it. A recipe request containing an allergy is fine. A long description of a symptom, with your child’s name and age, is a medical detail sent to a third party for an answer that will not be reliable anyway.

What “we use AI” tells you, and what it does not

Almost nothing on its own. The phrase currently covers everything from a median calculation to a full conversational model. Three questions get you past it:

• Which features use a model, and which are just maths? A good policy separates them.
• Is your data used to train anything? This should be a plain yes or no, and if it is a yes it should be optional.
• Which provider? An app that will not name the model provider is telling you something about the rest of its disclosures.

Advertising, brokers and the business model test

The most reliable privacy signal is not a promise. It is arithmetic: work out how the app gets paid. It sits alongside the other things worth weighing when choosing a baby tracker.

• Subscription — you are the customer. The incentives point the right way.
• One-off purchase — the same, though it can push development towards neglect rather than data.
• In-app advertising — ad networks receive an identifier and behavioural signals by design. That is how the ads get targeted.
• Free, no ads, no subscription, no obvious revenue — the question is worth asking directly. Sometimes the answer is a funded startup that has not decided yet, which is its own kind of uncertainty.

Pregnancy and baby data has a specific commercial value, because the arrival of a child is one of the few moments when household buying habits genuinely change. That is not paranoia; it is why the category attracts advertising interest at all.

Deletion is the one to test

Everything else is a claim about the future. Deletion is the thing you can check today, and it is worth checking before you have two years of history in the app.

• Open the settings and find out whether an account deletion option exists at all
• Check whether it deletes or merely deactivates — the wording usually gives it away
• Look for the retention period in the policy: how long after deletion do backups persist?
• Remember that uninstalling the app deletes nothing from anyone’s servers
• If deletion requires an email to support, expect it to be slow and to need chasing

Both major app stores now require a way to request account deletion for apps that support account creation, so its absence is not just inconvenient — it is out of step with the platform rules.

The rules that apply to children’s data

Two regimes matter most, and it is worth knowing what they do and do not guarantee.

• COPPA in the United States governs services collecting data from children under 13. Most parenting trackers are aimed at the parent as the account holder, which changes how it applies — the parent is the user, the child is the subject.
• GDPR in the UK and EU gives you rights over data regardless of who typed it: access, correction, deletion and portability, with children’s data treated as requiring particular care.

Be sceptical of badges. “COPPA compliant” and “GDPR compliant” are self-declared claims in most cases, not certifications issued by anyone. What is actually checkable is whether the policy names its processors, states a retention period, and gives you a working deletion route. Those are verifiable. A badge is not.

How Pixy answers each question

We publish Pixy, so treat this as disclosure rather than a recommendation — and check it against the privacy policy rather than taking it on trust.

• What leaves your phone. Account details, child profiles, growth data and activity logs, stored using Google Firebase so they sync across caregivers and devices. Offline logging is held on the device until you reconnect.
• Who else receives it. Three named services: Google Firebase for the database and authentication, RevenueCat for subscriptions, and Google Gemini for the AI assistants. That is the full list, and it is in the policy.
• Which features send data to a model. The AI Chef, AI Storyteller and AI Chatbot do. The Sleep Window prediction does not — it is a calculation over your own logs.
• Advertising. There is none, and Pixy does not sell user data. Premium subscriptions are how the app is funded.
• Deletion. Settings > Delete Account, inside the app, removing the account and its associated data. No email, no support ticket.

Where we are not able to make a claim, we do not: Pixy has not been through an external privacy audit or certification, and we would rather say that than display a badge that means nothing. Apply the same standard to every app you compare it against — including the ones with more badges.