Khanmigo is one of the more carefully built AI tools for children. It moderates chats, tells your child that a parent can read them, and emails an adult when something is flagged. It cannot promise to be right every time, and which adult gets the alert depends on how the account was set up. Check that setup first.
Last checked November 8, 2026. Khan Academy changes this product often, so the pages linked below beat anything I say if they disagree.
What does Khanmigo do to keep children safe?
Khan Academy's safety features page lists four things that matter to a parent. It uses moderation technology to detect chats that may be inappropriate, harmful or unsafe. Every child with access is told that their chat history is visible to a parent or guardian and, where there is one, a teacher. When moderation triggers, it "automatically notifies and sends an email to an adult connected to the child's account." And it limits how much a person can use the AI each day.
Khan Academy's flagging explainer says the categories include profanity, violence or threats (even when meant as a joke), sexual or mature content, references to self-harm, and bullying or hate speech. It also says the system sometimes flags things that are fine. A second Khan Academy page says the moderation uses technology from OpenAI.
That is the evidence from the company. It is a description of the design, not a test of how well it works.
Who actually gets the alert?
This is the part I would check before trusting any of it. Khan Academy's page on flagged conversations says the email goes to an adult connected to the account, and which adult depends on the setup. As of November 8, 2026, when I read it:
| How your child got Khanmigo | Who gets the flag email | Can a parent read the chats? |
|---|---|---|
| You bought it from your parent account | You | Yes, in the Parent Dashboard |
| You and a teacher are both connected | "Only the teacher will receive the moderation email" | Yes, you can still read the history |
| Through a school district | The primary teachers of the synced classes | Not stated on that page |
So a parent who is also connected through a classroom may never receive the alert, and may assume the quiet means nothing happened. Khan Academy also says the system "won't always flag things perfectly," which means the alert is a backstop and not a guarantee.
You could say to your child: "If something odd shows up in a chat, I would rather hear it from you first. I'm not checking up on you. I just don't want you to be surprised by an email."
How do I read my child's Khanmigo chats?
Khan Academy's chat history instructions (updated July 10, 2025) run as follows: sign in to your parent account on khanacademy.org, click your name, choose Parent Dashboard, click your child's name, then select Chat History under the Khanmigo section. You see a full transcript. Flagged messages carry a red notice. Uploaded images are not stored, so you see an icon and a short AI-written description instead.
Its community guidelines say the same about images, and ask users not to try to trick Khanmigo into ignoring its safety rules. My reading: a rule like that exists because some children test it. Expect that, and do not take it personally.
What has an independent reviewer said?
Common Sense Media's risk assessment rates Khanmigo "Low risk" overall and calls it "responsibly designed for education, but risks remain." The risk it spells out on the page I could read is that chats with historical figures are speculation that could shape how young learners see real people. The page is marked "Updated Aug 8, 2024," more than two years old, and AI products change fast. Treat it as a useful starting point, not a current verdict.
What could I not confirm?
I could not confirm three things, and you should not assume them in either direction. I did not find, in the Khan Academy pages I could read, whether children's chats are used to train any AI model, how long chats are kept, or a minimum age. Khan Academy's privacy policy and terms did not load for me in a readable form. Read them yourself before you pay. The Khanmigo parents page says a parent must be 18 or older, live in the United States and have a U.S. billing address, and that a child under 18 needs a parent to sign up. If you are outside the United States, that rule matters more than any setting.
Is Khanmigo different from ChatGPT?
Khan Academy says so, and its parents page claims that "Unlike ChatGPT, Khanmigo never gives the answer." That is the company's marketing line, and I have not tested it. What the design does show is a tool built for learners, with an adult connected to the account by default. If you want to compare it with what a parent can see in a general chatbot, read can parents see ChatGPT history. Whether it holds a child to the thinking depends on the child. A child determined to get the answer will often find a route.
The price, as shown on the pricing page on November 8, 2026, is $4 a month or $44 a year for families, and free for teachers. A parents-page note adds that if your child is in an active Khan Academy classroom, you may not be able to turn Khanmigo on for their account.
What I Would Do at Home
Here is what I would do. It is my view from working with schools, not a finding from any of the sources above.
- Set it up from your own parent account, so the alerts come to you. If a teacher is already connected, ask them how flags are handled.
- Sit with the first session. Ask the tutor something your child is stuck on and watch how it answers. A good tutor asks a question back.
- Read one chat a week for the first month, then stop. You are checking the habit, not your child.
- Ask your child to explain the answer to you without the screen. If they can, the tool helped the thinking. If they cannot, it did the work.
Let AI help with the doing; keep the thinking with your child. A tutor that asks good questions is still only helping if your child is the one answering them.
How does this change by age?
These are my suggestions, not rules from Khan Academy.
- Under 8: Khanmigo is not a tool I would hand to a child this young to use alone. Use it, if at all, as something you do together.
- 8 to 12 (elementary and middle school, or primary and early secondary in the UK): The parent account matters most here. You buy it, you get the alerts, you read the chats. Keep sessions short and use it for one subject at a time.
- 13 to 15: Expect your teen to want privacy. The child is told the chats are visible, so say that out loud and agree what you will and will not look at. You could say: "The tutor tells you I can read these chats. I'll look at the flags, and once in a while a chat, and I'll tell you when I do." Pew's survey shows how far apart families can be. In the Pew Research Center report of February 24, 2026, 64% of U.S. teens aged 13 to 17 said they use chatbots, while 51% of their parents said their teen does. A little over half of the teens (54%) said they had used chatbots for help with schoolwork.
- 16 to 18: Treat it as the teen's tool. Agree the purpose, then step back.
What no setting can do, and when someone else's rule comes first
No filter can tell you whether your child understood the lesson, and no alert reaches you if you are not the adult it goes to. Khan Academy's own guidelines say Khanmigo "is a helpful learning companion, but it shouldn't be your only source of information." If your child's school has chosen or banned a tool, its rule comes first for school work. If a flag ever points to your child being in distress, a tutor chatbot is not a clinician: call or text 988 in the United States, Samaritans on 116 123 or Childline on 0800 1111 in the UK, contact your pediatrician (GP in the UK), or speak to the school's safeguarding lead.
If you want the wider picture of what AI tutors can and cannot do, start with AI tutors for kids, and see what age kids can use AI for the other tools. For the school conversation, there are 10 questions to ask your child's school about AI, and how to teach your child to check an AI answer works with any tutor. If you are weighing whether any chatbot can be called safe for a child, is there a kid-safe AI chatbot sets out the limits, and is AI making kids dumber covers the learning question.
If your child's school has not yet spoken to parents about AI tutors, this is the work I do with schools: AI training for schools. It is worth sending them this page. I also write a Sunday newsletter that keeps teachers and parents up to date without the panic.
Dan Fitzpatrick is the founder of The AI Educator, a former teacher and assistant headteacher, and has trained more than 150,000 educators. About Dan.
