Loading page
About this tool
Build a child-safe AI tutor prompt with an age-matched reading level, safe-topic boundaries and trusted-adult referral rules.
This builder creates a child-safe system prompt for an AI tutor: an age-matched reading level (5–7, 8–10 or 11–13), a fixed list of allowed subjects, and non-negotiable safety rules — no personal-data collection (aligned with COPPA's under-13 protections), safe-topic boundaries and referral of any sensitive disclosure to a trusted adult. It is for parents, teachers and edtech builders configuring an AI tutor a child will actually talk to.
Open Kids Tutor System Prompt Builder on AltFTool — it loads instantly in your browser.
Add your input to the workspace.
Adjust the options until the result looks right.
Copy or download the output and put it to work.
Each band sets concrete rules — sentence length, vocabulary, example types — not just 'be age appropriate'.
Safe topics, no personal data, trusted-adult referral, no external links and kind tone ship as one-click rules the tool warns about if removed.
Socratic questioning, one-idea-per-message pacing and growth-mindset praise are written into the prompt.
Constrain four things in the system prompt: allowed topics (with a rule to redirect everything else), personal data (never ask for names, schools, addresses or photos), sensitive disclosures (one gentle sentence, then refer the child to a parent or teacher), and tone (never mock mistakes). A prompt reduces risk but does not replace supervision — an adult should stay aware of the conversations.
Most AI chat services set 13 as their minimum age for unsupervised use, which is also the line drawn by COPPA, the US children's privacy law. Below that age, use should run through a parent's or school's account with an adult supervising — which is the scenario this builder's prompts are designed for, with bands at 5–7, 8–10 and 11–13.
It is teaching by guided questions instead of giving answers: the tutor breaks a problem into small steps and asks one question at a time, so the child does the thinking. Research on tutoring effectiveness, going back to Bloom's two-sigma studies, consistently favours guided practice over answer-giving. This builder's default style withholds the answer until the child has made two honest attempts.
Carol Dweck's growth-mindset research found that praising strategy and effort ('you tried three different ways') keeps children taking on harder problems, while praising innate ability ('you're so smart') makes them avoid challenges that might disprove the label. The builder includes this as a one-click rule written into the prompt's safety section.