Before You Trust a Child-Friendly AI App - Five questions to ask before a learning tool, chatbot, or smart toy joins your family
A cheerful character, an educational label, and a promise to help with homework can make an AI app feel reassuring. The harder question is what sits behind that friendly introduction. Who receives your child's information? What happens when the conversation becomes personal? Who helps when the tool gets something wrong?
There is a useful new reason to ask. IEEE Standards Association has announced its first Responsible AI for Children product certification, awarded to the educational platform d-AI-logue. Its announcement describes independent assessment, privacy protections, age-appropriate design, and continuing accountability as products change. That is a welcome development, although an announcement about one certified product cannot establish the safety of every app marketed to families.
The practical lesson is to ask for evidence before granting access. Parents should not need a postgraduate qualification and three uninterrupted evenings to make a sensible decision. A small set of questions can help you decide whether to investigate further, supervise a limited activity, or leave a product alone.
Start with the job, and check the age rules
Ask what your child actually wants the tool to do. Explaining a science concept, practising vocabulary, and offering ongoing friendship are different jobs. Give the proposed activity a clear description, such as helping us generate questions about volcanoes. That makes it easier to judge whether the features and information requested are necessary.
Check the provider's current minimum age, account conditions, and parental permission requirements before creating an account. A colourful design is insufficient evidence of suitability. If your child is below the permitted age, choose another activity rather than changing their date of birth. Where the rules are unclear, ask the provider or school to explain them before proceeding.
For this age group, my suggested starting point is a specific, adult-supported learning task, where the service permits it. Treat a tool that encourages an ongoing personal relationship as a separate decision requiring much closer scrutiny. There is no obligation to introduce conversational AI simply because it is available, or because another family has already downloaded it.
Question one: what does the safety claim actually cover?
A claim about safety needs a subject, a scope, and some evidence. Ask which product was assessed, which version or features were included, who performed the assessment, and where you can check the result. A company's participation in a programme and a particular product receiving certification are different claims.
When a provider points to a badge, look for a direct explanation from the organisation behind it. Ask whether assessment covers the way your child would use the service, including voice conversations, uploads, and any additional features. If the explanation leaves those questions unanswered, keep them open rather than filling the gaps with optimistic assumptions.
These are suggested questions for evaluating a claim, not a description of any particular certification's full requirements. Even a meaningful assessment has boundaries. Use it as one input alongside your child's age, the intended activity, the provider's terms, and the controls you can actually operate.
Question two: what information must my child hand over?
Before the first conversation, look at what the service collects and what you can avoid providing. Ask whether prompts, voice recordings, photographs, and uploaded documents are stored, how long they are retained, and whether they may be used to train AI. Find out whether settings change those practices and whether deletion is explained clearly.
UNICEF's guidance on AI and children identifies privacy, safety, transparency, and accountability as requirements for systems that respect children's rights. Those principles offer a useful basis for questions to providers, even when a parent cannot inspect the technology itself.
Try a household rule: use invented details when real details are unnecessary. In a hypothetical homework activity, a child can ask for help planning a story about a fictional school without uploading a worksheet bearing their own name, class, and teacher's details. Read anything you plan to upload together, including information about other children.
Your child's curiosity does not require an all-access pass to family life. If a product cannot explain why it needs a photograph, location, or personal history for the activity you have chosen, pause. You can ask for an explanation, use a less revealing activity, or choose a different resource.
Question three: can a homework helper become a confidant?
The distinction matters because children may use tools beyond their advertised purpose. In an Australian survey of 1,950 children aged 10 to 17, conducted from February to March 2026, eSafety found that half of children who had used AI assistants had used them for at least one personal or social reason. The report also describes the increasingly blurred boundary between assistants and companions.
That finding is not a prevalence estimate for every child, and it does not show that all AI use causes harm. It does give parents a reason to ask what happens when a learning conversation turns towards friendship, worries, or personal advice. A tool's homework label cannot answer that question for you.
Explain the boundary in ordinary language: a chatbot produces responses, but a friendly response does not make it a person who knows you or takes responsibility for your wellbeing. Agree that worries about friendships, bodies, safety, or distress come to a trusted adult. You can practise naming that adult before there is a problem.
Question four: can we check an answer and get help?
Plan to verify important answers using an appropriate source, such as the assigned textbook or a teacher, rather than treating confident wording as proof. For a simple learning task, ask your child what they already know, what seems uncertain, and where you could check together. The goal is to make questioning part of using the tool.
Also find the reporting route before it is needed. Can you report a disturbing response from the conversation itself? Is there a human contact? Can you stop the interaction and remove an uploaded file? Try finding these controls as an adult, without asking your child to explore risky material to test them.
eSafety warns that AI companions can offer harmful content and dangerous advice. Its parent guidance recommends discussing online interactions and reassuring children that adults will help, including when they think they have done something wrong. Make that reassurance specific: telling me about an upsetting response will be the start of helping you.
Question five: what would make us change our decision?
Write down a few reasons to stop and review. Your list might include an unexpected request for personal information, content beyond the agreed activity, confusing purchases, or a new feature that changes how conversations work. These are proposed household boundaries, not evidence that every product behaves this way.
Revisit the decision when the service changes or your child wants to use it differently. An app you considered for shared vocabulary practice deserves a fresh look before it becomes a place for private conversations. Keep the review practical by checking the features involved in that change, rather than trying to anticipate every possible future use.
There are limits to what families can establish. Reading a policy cannot prove that a safeguard works, and supervision cannot compensate for every design failure. Providers and schools still need to explain their choices and demonstrate appropriate protections. Parents can reasonably expect understandable answers, rather than being left to translate reassuring slogans into a safety plan.
For now, choose one tool your child already encounters and work through these questions together. Keep the activity specific, personal information limited, and the route to a trusted adult obvious. A friendly interface may open the conversation, but clear evidence and workable boundaries should determine whether the app earns a place in your family.
Public source material
1. IEEE Standards Association, Responsible AI for Children: IEEE Delivers a Trusted Safety Formula to Help Protect Kids Online. Issuer announcement, 14 October 2026. https://www.businesswire.com/news/home/20261014367947/en/Responsible-AI-for-Children-IEEE-Delivers-a-Trusted-Safety-Formula-to-Help-Protect-Kids-Online
2. UNICEF Innocenti, Guidance on AI and children, version 3.0. December 2025. https://www.unicef.org/innocenti/reports/policy-guidance-ai-children
3. eSafety Commissioner, Talking to machines: Children’s experiences with AI assistants and companions. Report download labelled August 2026; web page has no visible publication date. Survey: 2 February to 4 March 2026, Australian children aged 10 to 17, n=1,950. https://www.esafety.gov.au/research/talking-to-machines-childrens-experiences-with-ai-assistants-and-companions
4. eSafety Commissioner, AI chatbots and companions: risks to children and young people. 18 February 2025. https://www.esafety.gov.au/newsroom/blogs/ai-chatbots-and-companions-risks-to-children-and-young-people