Your kid is already talking to an AI, and there’s a decent chance you’d never know which one. Lifehacker just went through the seven biggest AI chatbots for kids safety, line by line, and the differences are wild. One locks teens out completely. One hands them everything with a shrug. Guess which one your 13-year-old wants.
Every big chatbot, one table
Here’s the whole landscape in one place. Then we’ll go through what each control actually does.
| Chatbot | Minimum age | Parent controls | The catch |
|---|---|---|---|
| ChatGPT | 13 (consent under 18) | Account linking, time limits, sensitivity settings | Teens can unlink themselves |
| Gemini | 13 (under 13 via Family Link) | On/off switch, inherited family settings | Shallow, no chat visibility |
| Meta AI | 13 | Topic reports, self-harm alerts | Linking is optional for teens |
| Siri (Apple) | 13 | Screen Time, feature toggles | Teens 13+ don’t need supervision |
| Copilot | 13 | On/off, time windows, feature blocks | Parents can’t read chats |
| Grok | 13 (on paper) | Kids Mode, self-declared | Age checks fail, per Common Sense Media |
| Claude | 18 | None needed | Adults-only by design |
A pattern jumps out fast. Almost everyone says 13. Almost nobody can actually verify it. So before trusting any of these AI chatbots for kids, look at what the controls really cover.
ChatGPT: the most complete setup
OpenAI has the deepest parental toolkit of the bunch, and it starts automatically. Users under 18 get moved into a “ChatGPT for Teens” experience whether they ask for it or not: a study mode for schoolwork, plus built-in blocks on sexual or romantic roleplay, graphic violence, and self-harm content. Age checks lean on behavioral signals, like what times of day someone uses the app and what they talk about, because nobody checks an ID at signup.
The parent side works through account linking. From the web app, you click your profile icon, open Settings, then Parental controls, then Add family member. Once linked, you can set quiet hours when the app goes dark and dial down sensitivity settings. You can’t read your kid’s chats, though. And here’s the part every parent should know before arguing about it: the teen can break the link. OpenAI says so itself.
One more loophole to note. Anyone can open a browser, go to ChatGPT’s site, and chat without an account. Those anonymous chats default to under-18 protections but can’t be saved, and there’s no age gate at all.
Gemini: simple but shallow
Google ties Gemini to Family Link, its existing family management system. If your child is under 13, they can only use Gemini through an account you manage, and you’re the one who flips the switch on. One button, clear rules. For under-18s generally, Google adds protections on top: the AI won’t roleplay a persona, it actively discourages emotional dependence, and it steers away from chats about drugs, bullying, and self-harm.
That’s the good news. The catch is depth. Gemini’s controls are basically on, off, and whatever settings your family group already has. You get no insight into conversations. And if your teen has their own unsupervised Google account, which most teenagers do, none of this applies. You can manage family settings through Google’s Family Link portal, but manage expectations first.
Meta AI: reports without transcripts
Meta AI lives where kids already are: WhatsApp, Instagram, and its own app. Teen accounts get automatic blocks on sexual roleplay and substance talk, and chats about eating disorders or self-harm trigger support resources instead of conversation.
The parental angle is different here. Through the Meta Family Center, parents get reports on the general topics their kid discusses with the AI. No transcripts, no exact prompts. If a chat touches self-harm or suicide, the parent gets an alert. Sounds useful, right? It is, with one asterisk: connecting a parent account is optional for teens. The safeguards apply either way, but the reporting only works if the link exists.
Siri and Copilot: the middle ground
Apple and Microsoft both run the same basic play: age 13 minimum, ecosystem-level controls, optional supervision.
Siri’s AI features need an Apple account, and kids under 13 can only get the older, basic Siri through an account an adult creates. For teens, Apple’s iOS 27 parental controls let you disable specific features like writing tools or image generation, but only through Family Sharing, and only if you got your kid into the family group in the first place. Screen Time does the heavy lifting for time limits.
Copilot mirrors that structure through Microsoft Family Safety. Parents can switch Copilot off entirely, set allowed time windows, and block features like voice mode or image generation. Chats stay private from parents. Microsoft adds one nice touch: users flagged as under 18 stop seeing personalized ads and results.
Grok: the one to skip
Now for the uncomfortable section. Grok, available through X and its own app, ranked worst among AI chatbots for kids in a recent Common Sense Media report, and the details back it up. Age checks don’t work, according to the report. The platform combines weak guardrails with content that includes explicit material and mental health risks, and it has a history with deepfake imagery that nobody’s proud of.
There is a Grok Kids Mode. It limits topics and content, sounds responsible in a screenshot, and falls apart in practice because it runs on self-declared age with no real verification. A kid types a different birth year and it’s gone. Common Sense Media’s advice is blunt, and I’d echo it: don’t let your teens use this one. Nothing on the roadmap suggests that’s changing soon. (If you want the flip side, our guide to spotting AI scams covers the same trust-me-bro pattern from the other direction.)
Claude: strict by design
Anthropic took the opposite path from everyone else. Claude requires users to be 18 or older, full stop. There’s no free anonymous web access like ChatGPT and Gemini offer, so there’s no side door either. If behavioral signals suggest a user is actually a minor, Anthropic can demand age verification, including a face age estimate.
Could a determined 16-year-old still get through? Probably, eventually. But Claude is the only major chatbot where the default answer to “should my kid use this” is a flat no from the company itself, which makes the parent conversation refreshingly short.
Set this up tonight
Enough comparison. Here’s the actual to-do list, and it takes about 20 minutes. Parental controls on AI chatbots for kids only work if someone turns them on, and that someone is you.
- 1. Ask your kid which AI tools they use. Expect an incomplete answer, then check installed apps together.
- 2. Link your ChatGPT accounts if they use it: profile icon, Settings, Parental controls. Set quiet hours for school nights.
- 3. For under-13s on Google, set up Family Link and flip Gemini on deliberately, never by accident.
- 4. Turn off Meta AI one-on-one chats where the setting exists, and connect the Family Center link if your teen agrees.
- 5. One conversation matters more than any setting: which topics are never okay to discuss with a chatbot, and who they come to when something feels wrong.
That last step isn’t a software feature. It’s the control none of these companies can install for you, and it’s the one that works at 11pm on a school night.
The takeaway
The parental controls on AI chatbots for kids are real but leaky: ChatGPT gives you the most levers, Gemini gives you a clean on-off switch, Meta tells you topics without showing transcripts, and Grok deserves the skip. Spend 20 minutes tonight on the account links, then spend five on the conversation, because the settings help but the talking is what actually protects them. If you want a second layer, our guide to ChatGPT’s locked-chats PIN feature shows how to keep chats private on shared devices.