Skip to content

    Is an AI tutor safe for children? Seven things to check

    Facts and prices checked on 26 September 2026.

    There is no single answer to whether an AI tutor is safe, because products differ a great deal. What you can do is check a handful of things before your child uses one. Here they are, with what the evidence and the Department for Education's 2026 standards say, and an honest account of where WonderPath stands today.

    Why the questions matter

    A widely cited study of nearly 1,000 secondary pupils (published in PNAS in 2025) found that pupils given a chatbot that gave answers on demand did 48% better during practice and 17% worse on the later exam without it. When the same kind of model was set up to give hints instead of answers, the harm disappeared. The lesson for parents is that how the helper behaves matters more than the fact that it uses AI.

    The DfE's Generative AI product safety standards were updated on 19 January 2026. They are guidance for suppliers to schools, not law, but they are the clearest public checklist. Among other things they ask products to "only show a full solution after a genuine learner attempt", to follow "a pattern of progressive disclosure of information", to "detect cognitive offloading", and to "include default time limits on usage".

    Seven things to check, and where WonderPath stands

    Seven checks for an AI tutor and WonderPath's current position
    CheckWhy it mattersWonderPath today
    Does it give hints or answers?Answers on demand can raise practice scores while lowering what a child can do alone.In quizzes and practice: hints first, and the answer only after the quiz. The chat helper works the same way and declines requests for the final answer.
    Can you read every conversation?You cannot supervise what you cannot see.Yes: every conversation is in your account, and you can switch the helper off for any child.
    Is it a person or a companion?The DfE standards ask products not to give the impression of a person or a friend.It says it is a computer program if asked and does not claim feelings. It uses friendly subject names, which we are reviewing against the DfE standard.
    What checks run on messages?Children will test the limits.Messages are checked against our blocked-topic list and an automated moderation service, and unsafe topics are refused and recorded. No filter is perfect.
    Where does the data go?Children's data needs the highest care.Never sold and never used for advertising. We use AI services whose terms exclude training on our data, and we are confirming that setting with each provider. The processors are named in our privacy policy, with where each one stores data.
    Who checks the lessons?AI can be wrong.Lessons are written by AI. Review by a qualified teacher is being set up and is not in place yet, so please treat lessons as drafts.
    Does it try to keep children hooked?The DfE standards name flattery, guilt and streak pressure as manipulation.A cap of 30 helper messages a day, break prompts, no adverts, and a streak that costs nothing to miss.

    Questions to ask any provider

    1. Does the helper give the answer, or make my child try first?
    2. Can I read every conversation, and switch it off?
    3. What happens if its safety checks fail or are unavailable?
    4. Which companies process my child's data, and in which countries?
    5. Who has checked the lessons, and can I see how many have been checked?

    What the rules say today

    Ofcom's published guidance treats a chatbot that only talks one-to-one with a user, with no sharing between users, as outside the Online Safety Act at present. The government has said it intends to tighten the rules for chatbots used by children, so check gov.uk for the latest. UNESCO's 2023 guidance suggests a minimum age of 13 for generative AI in classrooms. For children under seven we recommend an adult sits alongside.

    Our full position, including what is still being built, is in the AI safety policy and on the Beta page.

    Common questions

    Is an AI tutor safe for a 6 year old?

    It depends on the product and on how it is used. We recommend an adult sits alongside a child under seven, reads the conversations, and switches the helper off if in doubt. You can turn WonderPath's helper off per child.

    Can an AI tutor replace a teacher or tutor?

    No. It can give practice and explanations at any hour, but it can make mistakes and does not know your child as a person. Use it alongside school, not instead of it.

    What do the DfE standards say about AI tutors?

    They ask products to withhold full solutions until a learner has tried, reveal help step by step, detect when a learner is asking the system to do the work, avoid giving the impression of a person, and avoid manipulation. They are guidance for suppliers to schools, not law.

    Try WonderPath free

    One child, daily lessons and an AI helper you can read and switch off. No card needed.

    Start free

    Keep reading

    Sources

    How this page was made: written by the WonderPath team with AI assistance and checked against the sources named above on 26 September 2026. Prices and features change; if anything is out of date, tell us at info@wonderpath.co.uk. WonderPath is a competitor of some products named here, and we have tried to describe them fairly.

    A note for parents: We use cookies to understand how families use WonderPath so we can make it better. No personal data about your child is collected. Cookie policy