Chatbots feel like private text threads — but privacy depends on the app, the account, and the settings. Here's what your child should never type into any of them, and why.

Kids have grown up texting, and a chatbot looks and feels like a text thread: you type, it answers, it remembers. That familiarity is the problem. AI is no longer one app — teens meet it in homework helpers, in search, in Snapchat's My AI, in character and companion apps, in school tools, and inside the social apps they already use.

And teens are using it heavily: Pew found that 64% of U.S. teens use AI chatbots, including about three in ten who use them daily. The oversharing is already real — Common Sense Media found that a quarter of teens have shared personal information with AI companion platforms. Not because kids are careless, but because these tools are built to feel like a private, trusted listener.

When you type into a consumer chatbot, your words don't vanish after the reply. On many consumer services the default is that conversations may be used to help train future models unless you find and switch off a data-controls setting — and some services say conversations may be reviewed by people in certain cases, such as safety, quality, or product improvement. (This isn't random employees reading every chat; it's documented review for specific purposes — check each service's policy.)

Crucially, the rules are not the same everywhere. Privacy, retention, human review, age limits, and whether your data trains the model differ by company, by product, and by account type. OpenAI's ChatGPT, Google's Gemini, Microsoft's Copilot, Snapchat's My AI, and Anthropic's Claude each document their own behavior — and a free consumer account is treated differently from a school, enterprise, or supervised child account.

Two facts make "I'll just delete it later" an unreliable safety plan. First, retention rules vary by app — Snapchat, for example, says content shared with My AI is kept until you delete it, and that My AI can read group-chat messages when it's included. Second, bugs happen: in one documented 2023 incident, a software bug briefly exposed some ChatGPT users' chat-history titles to other users. None of this means AI is unsafe to use. It means the safest privacy habit is not to type sensitive information in the first place.

Put this on the fridge. The simple version of the rule: don't type anything that would identify, locate, expose, or embarrass you or someone else. Specifically, none of these belong in a chatbot:

You don't need your child to memorize seven bullet points. You need one habit. Say this to them:

If they wouldn't put it on a public bulletin board at school, it doesn't go into an AI.

A personal free account, a school-provided account, a supervised child account, and an enterprise or work account may have very different privacy rules. Before your child uses AI for school, ask four questions: who provides the account; whether chats are used to train the model; how long data is retained; and whether a teacher or admin can review activity. The same chatbot can be a different privacy product depending on which door your child walks in through.

Ages 3–6. Children this young shouldn't use open chatbots unsupervised, and AI now appears inside toys and kids' apps. If you're using AI near them, model the behavior out loud: "I'm not going to type our address in here."

Ages 7–10. Do not give children this age unsupervised access to open-ended chatbots. Introduce the bulletin-board rule concretely, and keep AI use in shared spaces while you review the checklist together. Where you do allow it, supervised child experiences exist — for example, Google offers supervised access to Gemini for children.

Ages 11–13. Many kids begin using AI more independently here — often through Snapchat's My AI or character apps rather than a standalone chatbot. Most services set a minimum age around 13 and offer parental controls; walk through the checklist once and make it a standing rule.

Ages 14–18. Teens are the heaviest users and the least likely to mention it. Give them the real reasons — training data, human review, retention that varies by app, and breaches — and frame it as digital self-defense: what they type could resurface, and they own the habit of protecting it.

[DOWNLOAD] 📎 Free download: "The Never-Share Checklist" — the fridge-ready one-pager with the rule, the list, and the four-question habit.

If your child is using a chatbot to talk about self-harm or a crisis, a bot is not a safe substitute for help. If your child may be at immediate risk of self-harm or suicide, call emergency services or contact a crisis professional right away. In the U.S., the 988 Suicide & Crisis Lifeline is available 24/7 by calling or texting 988. In Canada, call or text 988; Kids Help Phone is also available at 1-800-668-6868 or by texting CONNECT to 686868. This article is general guidance for parents and is not medical advice.

Disclosure: Hossein works in AI and builds AI-related products. AI by Age takes no AI-vendor sponsorships. Full disclosure on our About page.