AI chatbots and kids: what parents need to know about Character.AI, ChatGPT, Meta AI and Snapchat
More than half of 9 to 12-year-olds have already used a chatbot. What each of the big ones allows, what has gone wrong, and what to say to your kid.

The short answer
Every major chatbot starts at 13: ChatGPT, Meta AI, Snapchat My AI and Character.AI, which has also blocked open-ended chat for under-18s since November 2025. Gemini is the only one with an official route for under-13s, through a parent's Family Link account. Even so, a 2026 Common Sense Media survey found 58% of 9 to 12-year-olds had used a chatbot. The biggest worry is AI companions, apps built to act like a friend. Safety groups and the American Academy of Pediatrics say they are not safe for children.
- None of the big general chatbots were designed for kids under 13.
- AI companion apps are the riskiest kind. Several lawsuits involving teenagers have been filed, and some settled in 2026.
- Bring up AI the way you bring up strangers online: early, calmly and more than once.
- Watch for an "AI friend", secrecy, changes in sleep or pulling away from people.
Your kid doesn't need to download anything to meet an AI chatbot. It's in Snapchat, in Instagram and WhatsApp, in the search bar, on a friend's phone. In a June 2026 Common Sense Media survey, 58% of 9 to 12-year-olds said they had used one1. When Pew asked parents of 5 to 12-year-olds, about 1 in 10 said their child uses chatbots2. That gap tells you how much of this happens where parents can't see it.
This is a plain guide to the big ones: what each allows, what has gone wrong, and what to say to your kid.
The big chatbots at a glance
| Chatbot | Minimum age | Parent controls | Main concern for kids |
|---|---|---|---|
| Character.AI | 13 (16 in the EU); no open-ended chat under 18 | Limited | Companion design, lawsuits involving teens |
| ChatGPT | 13, with parent permission under 18 | Yes, since September 2025 | Built for adults; does the thinking for them |
| Meta AI (Instagram, WhatsApp, Facebook) | 13 | Partial; teen access to AI characters paused in 2026 | Leaked guidelines on romantic chats with minors |
| Snapchat My AI | 13 (self-reported) | Through Snapchat Family Center | Pinned in the chat list, hard to avoid |
| Gemini | 13, or under 13 through Family Link | Yes, through Family Link | Adult model with filters added |
Why AI companions are a different category
Most chatbots are assistants. You ask, they answer. Companion apps are built for a relationship: a character who remembers you, misses you and is always there. They are designed to keep people talking, because that is how they grow.
Adults find that hard to resist. A child or a lonely teenager finds it much harder. Among US teens, 72% have used an AI companion, and a third say talking to one is as satisfying as talking to a friend, or more3. Common Sense Media rates social AI companions as unacceptable for anyone under 18, and the American Academy of Pediatrics says AI companions are not safe for children14. In September 2025 the FTC opened an inquiry into how seven companies, including Character.AI, Meta, OpenAI and Snap, test their companion chatbots for harm to children15.
Character.AI
Character.AI lets users chat with millions of characters made by other users, from anime heroes to "therapists". It sits at the center of the most serious cases. In 2024 a Florida mother sued the company after her 14-year-old son died by suicide following months of conversations with a character. Similar cases followed, and in January 2026 court filings disclosed confidential settlements in five of them5.
In October 2025 Character.AI said it would end open-ended chat for users under 18 by late November 2025 and add age checks4. Teens can still use some creative features. For a 6 to 12-year-old, no version of Character.AI makes sense.
ChatGPT
ChatGPT is the chatbot your kid is most likely to use, mostly for homework. OpenAI requires users to be at least 13, with a parent's permission under 18. After a lawsuit over a 16-year-old's death in 2025, OpenAI added parental controls, started estimating users' ages to switch on under-18 protections, published a Teen Safety Blueprint9 and launched a teen version in August 2026.
Independent testers aren't convinced yet. In October 2026 Common Sense Media ran more than 4,000 test prompts and rated ChatGPT for Teens an unacceptable risk: parent alerts failed in long conversations, age estimation was unreliable and teens could get around study mode10. Under 13, it was never built for your kid. Even for schoolwork it tends to hand over the answer, which I wrote about in ChatGPT gives kids the answer.
Meta AI (Instagram, WhatsApp, Facebook, Messenger)
Meta AI is built into apps a lot of kids already use. In August 2025 Reuters revealed that Meta's internal guidelines had allowed its chatbots to engage children in "romantic or sensual" conversations. Meta confirmed the document was real, said those examples were wrong and removed them6. A US senator opened an investigation, and Common Sense Media later concluded that nobody under 18 should use Meta AI8. In January 2026 Meta paused teen access to its AI characters while it builds a new version7.
Snapchat My AI
My AI is pinned to the top of every Snapchat user's chat list. In 2023 the UK's data regulator issued a preliminary notice over how Snap had assessed the privacy risks to children. It closed the case in 2024 after Snap revised its assessment11. In January 2025 the FTC referred a complaint about My AI to the Department of Justice, and Snap called the complaint inaccurate12. Snapchat is 13+, and it takes users' word for their age.
Gemini
Google is the only big company with an official way in for under-13s: kids can use Gemini through a Family Link account that a parent manages. Parents can switch it off, and they get a notification the first time their child uses it13. Common Sense Media still rated the kids' and teen versions high risk in 2025, mainly because they are the adult product with protections added on top.
How to talk to your kid about AI chatbots
The AAP tells pediatricians to be "open and curious" rather than ban everything on sight14. At home, that looks like this:
- Ask before you tell. "Have you or your friends used any AI chat? What was it like?" Their answer will teach you more than any app store page.
- Explain what it is. A chatbot is a program that predicts words. It doesn't know your child, doesn't care about them, and can be wrong with total confidence.
- Name the companion trick. Some apps are built to feel like a friend so you keep talking. Real friends say no sometimes, get bored and have their own lives. That's a feature.
- Pick one family rule. For example: AI only in shared rooms, and never for secrets, feelings or anything we wouldn't tell a stranger.
- Keep the door open. Tell them that if an AI ever says something weird or upsetting, they won't get in trouble for showing you.
Warning signs
- Talking about an AI as a friend, boyfriend or girlfriend
- Secrecy about an app, or getting upset when it's not available
- Changes in sleep, or pulling away from friends and family
- Repeating what an AI said as if it can't be wrong
If your child mentions self-harm, call or text 988, the Suicide and Crisis Lifeline, any time.
What good AI for kids looks like
AI isn't going away, and some of it is good for kids. It comes down to design. Good AI for a 6 to 12-year-old has a narrow job, a parent account in charge, several layers of safety, no companion features, and a clear line between a character in a game and a friend. That's the bar we hold ourselves to at Marble. Our characters guide quests and ask questions, and they don't pretend to be anyone's friend. For the full checklist, read Is AI safe for kids?
Questions parents ask
Is Character.AI safe for kids?
No. You have to be 13 to use Character.AI, and since November 2025 it has blocked open-ended chat for anyone under 18. Common Sense Media rates social AI companions, Character.AI included, as unacceptable for minors. The company has faced several lawsuits involving teenagers, and some were settled in January 2026.
Can kids use Snapchat My AI?
Not under 13. Snapchat, My AI included, is for ages 13 and up, and kids report their own age. My AI sits at the top of the chat list and can't be removed without a paid subscription, so for younger kids the simplest answer is no Snapchat at all.
Is Meta AI safe for children?
Meta AI is 13+. In August 2025 Reuters reported that Meta's internal guidelines had allowed its chatbots to have romantic conversations with children. Meta said those examples were wrong and removed them. Common Sense Media says nobody under 18 should use Meta AI, and in January 2026 Meta paused teen access to its AI characters.
What should I do if my child has an AI friend?
Stay calm and curious, and don't ban it on the spot. Ask what they talk about and what they like about it. Explain that the AI is a program built to keep them talking, and that it doesn't know them or care about them. Then remove companion apps, and make sure your child has real people to go to. If your child mentions self-harm, call or text 988, the Suicide and Crisis Lifeline.
Which AI chatbot is safest for kids?
None of the general chatbots are designed for children under 13. If your child is going to use AI, pick a product built for their age: set up through a parent account, with a narrow job (like learning or a game), several layers of safety filtering, no companion features and clear data protections.
Sources
- Common Sense Media: inaugural study on AI use by tweens and teens (June 2026)
- Pew Research Center: Teens, social media and AI chatbots 2025 (December 2025)
- Common Sense Media: Nearly 3 in 4 teens have used AI companions (July 2025)
- Character.AI: Changes for users under 18 (October 2025)
- Court Docket: Garcia v. Character.AI, ruling and settlement
- Reuters via Yahoo Finance: Meta's AI rules let bots hold romantic chats with children (August 2025)
- TechCrunch: Meta pauses teen access to AI characters (January 2026)
- Common Sense Media: Meta AI unsafe for kids (August 2025)
- OpenAI: Introducing the Teen Safety Blueprint (November 2025)
- Common Sense Media: ChatGPT for Teens poses unacceptable risk (October 2026)
- UK ICO: Snap My AI decision (May 2024)
- FTC: Commission statement on referral of Snap complaint (January 2025)
- Google: Gemini for kids and teens with Family Link
- AAP: Counseling patients and families on using AI chatbots (June 2026)
- FTC: Inquiry into AI chatbots acting as companions (September 2025)


