The central rule: increase independence slowly and supervision quickly. A low-stakes creativity activity is different from a conversation about health, relationships, private images, money, or personal safety.
Age is a guide, not a safety guarantee
Development varies. A learner may explain a technical concept well and still need help recognizing manipulation, protecting another person’s information, or stopping an upsetting interaction. Consider four things together: age, maturity, the design of the tool, and the consequences of a wrong answer.
Ages 8–10: build the mental model
At this stage, focus on what AI does rather than giving unrestricted access to a chatbot. Use adult-operated demonstrations, paper activities, sorting games, and familiar examples such as recommendations or autocomplete.
Useful goals
- Explain that AI finds and generates patterns; it is not a person or source of guaranteed truth.
- Recognize that confident language can still be wrong.
- Keep real names, passwords, school details, locations, and private images out of a prompt.
- Ask an adult before using a new AI feature.
Try this
Write three short statements: one verified fact, one opinion, and one invented claim. Ask the learner what evidence would help distinguish them. The activity teaches that the appearance of a sentence does not prove its reliability.
Ages 11–12: practise checking and stopping
Learners can begin comparing outputs and sources under close supervision. Keep accounts and free-form connected tools adult-managed, especially when a service can retain input or respond unpredictably.
Useful goals
- Find the specific claim that needs verification.
- Compare an AI answer with a current primary or teacher-approved source.
- Recognize invented citations and check that a source actually exists.
- Stop when an output becomes private, sexual, threatening, hateful, or emotionally serious.
Try this
Give the learner a fictional answer containing one wrong date. Ask them to underline the checkable claims, select a source, and record what changed after verification.
Ages 13–15: connect AI use to responsibility
Teenagers may use AI for school, creativity, coding, or exploration. The emphasis should shift from simple permission to visible responsibility: following school rules, protecting other people’s information, checking claims, and explaining the human contribution.
Useful goals
- Understand the difference between brainstorming, editing, translation, generation, and submitting AI-created work.
- Follow school disclosure and citation requirements.
- Recognize bias, stereotypes, persuasive design, and synthetic media.
- Avoid treating a chatbot as a confidential friend, therapist, or sole source of advice.
Try this
Compare two versions of a paragraph: one written independently and one revised with AI. Ask the learner to mark every change, verify factual additions, and write a short authorship statement.
Ages 16–17: prepare for independent use
Older teenagers benefit from practice with real workflows and explicit limits. They should be able to decide when AI adds value, when it creates unacceptable risk, and how to document the process for school or work.
Useful goals
- Evaluate provider terms, privacy settings, retention, training-use choices, and age requirements before using a tool.
- Choose appropriate sources for current, scientific, legal, health, or financial claims.
- Understand copyright, confidentiality, consent, and reputational risk.
- Escalate high-stakes decisions to a qualified person.
Try this
Have the learner write a one-page tool assessment: purpose, data entered, provider, retention, important risks, verification plan, and when they would choose not to use it.
Supervision should follow the stakes
Generating fictional character ideas, creating practice questions from material already understood, or asking for another explanation. Continue protecting privacy and checking factual claims.
School assignments, persuasive writing, news, coding, or content about another person. Follow the relevant rules, disclose assistance, and review every consequential output.
Health, law, money, sexuality, self-harm, abuse, threats, private images, emotional dependence, or personal safety. Move the conversation to a trusted and qualified human.
A five-question parent check
- Does the tool allow free-form text, images, voice, or uploads?
- What information leaves the device, who receives it, and how long is it kept?
- Can the learner block, report, delete, and ask for adult help?
- Does the activity match the learner’s maturity and the consequence of a mistake?
- Can the learner explain what the AI did, what the human did, and what was verified?
Do not use age alone as consent. A birthday field or parental gate may help route a user, but it does not automatically satisfy children’s privacy requirements. Review the tool’s current terms and privacy practices.