Handing a child an AI tool with no setup is like handing them a car with no seatbelt and no driving lesson. It might be fine. But five minutes of setup removes most of the risk and turns a vague worry into a calm, supervised routine. Here's the whole process, in order.
- Set it up as the adult, on your account, with a real age entered — most assistants require 13+.
- Turn off chat history/training where you can, and teach the golden rule: no personal details in prompts.
- Allow mainstream assistants (with supervision) and kid-focused tools; avoid open-ended companion chatbots.
- Agree the rules before the first login, then do session one together. The conversation matters more than any setting.
The 60-second version
If you only do four things: (1) you create and own the account; (2) you enter the correct age and check the tool allows it; (3) you switch off history/training and teach "no personal info"; (4) you agree simple rules and sit in on the first session. Everything below is the detailed version.
Step 1: Check the age rules
Most general AI assistants set a minimum of 13, and some require parental consent under 18. Under-13s should use kid-focused tools, with these assistants operated by a parent rather than the child. Don't fake the birthday — the age rule exists partly because default settings and content aren't tuned for younger children.
Step 2: Set up the account (as the adult)
- Create the account with your email, and keep the login. The account is yours; your child is a supervised user on it.
- Use a strong, unique password. Turn on two-factor authentication.
- If the tool offers a teen or supervised experience (some do), use it.
Step 3: Lock down privacy
- Turn off chat history / "improve the model" / training where the setting exists, so conversations aren't used to train the AI.
- Review data controls and delete history periodically.
- Teach the golden rule out loud: full name, school, address, phone, passwords and photos never go into a prompt. Make it a habit, not a one-off lecture.
Kids share without meaning to — "help me write to my teacher Mrs Advani at Oakridge School" just leaked a name and a location. Practise swapping real details for placeholders: "my teacher," "my school."
Step 4: Choose the right tools
Not all AI is the same category of risk. A rough guide:
| Allow (with supervision) | Approach with caution | Avoid for kids |
|---|---|---|
| Mainstream assistants (13+): ChatGPT, Gemini, Claude | Image generators (adult sets up; check outputs) | Open-ended companion chatbots (Character.AI, Replika) |
| Kid-focused, education-first AI tools | Anything with an in-app social/community feed | "AI girlfriend/boyfriend" or roleplay-companion apps |
| School-provided, vetted tools | New apps you haven't checked yet | Tools with no clear age policy or privacy stance |
For why the companion category is different, see our guide to AI safety for children and our deep dive on spotting fakes. To compare the mainstream assistants, read our guide on ChatGPT for kids.
Step 5: Agree the rules first
Settings are the easy half. The durable protection is a short, shared agreement made *before* the first login — where, when, what for, and what to do when something feels off. Rather than reinvent it, use our ready-made conversation starters and pair this setup with a written family agreement (we have a free template).
- Where & when: shared spaces, agreed times — not overnight, not the bedroom.
- What for: learning, creating, help with tasks — not secrets, not companionship.
- The honesty line: AI can coach you; it can't do your work for you.
- The 'tell me' rule: if anything is scary, weird or upsetting, come to me — you won't be in trouble.
Step 6: Do the first session together
Sit beside your child for the first go. Show them a good prompt, point out when the AI sounds confident but is actually guessing, and celebrate a moment of catching a mistake. Fifteen minutes of shared use teaches more than any settings page — and it signals that AI is something you do *with* them, not something you police from a distance.
AI Fundamentals for Kids teaches 10–16 year olds how AI works, how to prompt it, how to check it and how to stay safe — with a full module on safety, privacy and honesty. Self-paced, no coding, and you can preview every module yourself.
Explore the course →Ages 10–16 · No coding · Parent-guided · 7-day money-back guaranteeFrequently asked questions
What age can a child use ChatGPT, Gemini or Claude?
Most general assistants set a minimum age of 13, sometimes with parental consent required under 18. Under-13s should use kid-focused tools, with any mainstream assistant operated by a parent. Always check the current terms, as they change.
How do I make ChatGPT safe for my child?
Set it up on your own account with the correct age, turn off chat history/training in settings, teach the rule that no personal information goes into prompts, agree clear rules, and do the first session together. Supervision plus a short agreement does more than any single setting.
Are there parental controls for AI?
Some tools offer teen or supervised experiences and limited activity insights, but AI parental controls are less mature than for social media. The most reliable controls are adult-owned accounts, privacy settings switched on, tool choice, and an open agreement — not a single toggle.
Which AI tools should I avoid for my child?
Open-ended companion chatbots (like Character.AI or Replika) and 'AI friend/partner' apps, because they're designed for emotional attachment and are rated unsafe for minors. Also avoid any tool with no clear age or privacy policy.
Do I need to sit with my child every time they use AI?
Not every time. Sit in for the first session of any new tool and for younger children; older, well-briefed teens can work independently and come to you with what they made or with anything odd. Build independence on top of a good setup.