ChatGPT is not built for children under 13, and OpenAI requires parental permission for anyone aged 13 to 18. ChatGPT's parental controls let you set quiet hours, turn off image generation, opt your teen out of model training, and receive a safety alert if the system detects serious self-harm risk. They do not let you read your teen's conversations. That single gap surprises most parents, and it is the difference between a setting that protects your teen and a setting that just makes you feel like it does.
Is ChatGPT safe for kids? The short answer by age
OpenAI draws hard age lines that parents often miss. Under 13, the answer is no: the product is not meant for them and they are not permitted to have an account. From 13 to 17, a teen may use it only with a parent's permission, and OpenAI now applies extra content protections to accounts it believes belong to minors. At 18, your child becomes an adult user with full access and no parental layer at all. So the honest answer is not a flat yes or no. It is a yes for a supervised teen on the right settings, a no for a young child, and a moving target in the years between as your child earns more independence.
| Age band | OpenAI's position | What you should do |
|---|---|---|
| Under 13 | Not permitted. ChatGPT is not designed for this age and accounts are not allowed. | Use a kid-specific tool with adult supervision, or wait. Do not create a teen account by faking a birth year. |
| 13 to 15 | Allowed with parental permission. Higher exposure risk; less context for handling adult or distressing replies. | Link accounts, turn on quiet hours, disable image generation and memory, and opt out of training. Sit with them for the first sessions. |
| 16 to 17 | Allowed with parental permission. Minor protections still apply to accounts OpenAI predicts are under 18. | Keep accounts linked for safety alerts, but lean on conversation rather than restriction. Discuss accuracy, sources, and what to never share. |
| 18+ | Full adult account. No parental controls, no minor content layer. | Have the privacy and verification talk before the birthday, because the guardrails disappear the moment they turn 18. |
What ChatGPT's parental controls actually do
The controls cover settings and schedules, not surveillance. After you and your teen link accounts, you manage a defined set of options from your own login. Most of the safety value sits in two or three of these toggles rather than spread evenly across all of them, so it pays to know which ones carry the weight.
- Quiet hours: set a time window when ChatGPT cannot be used from the linked teen account, useful for school nights and bedtime. Only one window can be active at a time.
- Turn off voice mode, so the account is text only.
- Disable memory, so ChatGPT stops carrying details across conversations.
- Remove image generation, blocking the account from creating pictures.
- Opt the teen out of model training, so their content is not used to improve OpenAI's models.
- Reduce sensitive content by default, including limits on graphic material, viral challenges, sexual or romantic roleplay, and extreme beauty ideals.
- Safety notifications: in limited cases where the system flags possible signs of serious self-harm, you may receive an alert with only the information needed to support your teen.
Linking is consent-based on both sides. A parent sends an invite; the teen must accept it before any control applies. You cannot quietly attach to a teen's existing account, and the teen can see that controls are on.
The surprise: you cannot read the chats
Parental controls do not let you read or monitor your teen's conversations. OpenAI built it that way on purpose, and here is what most parent guides will not tell you: the transcript stays private to your teen no matter which toggles you flip. You can shape the experience and shut it off during set hours, but you never see the messages. The only exception is narrow. If OpenAI's systems detect signs of serious safety risk, OpenAI may notify a parent, and even then only with the minimum information needed to help. Notification depends on the severity of what was found, so a passing dark joke is treated differently from a sustained pattern that suggests real danger.
Read every message and you do not make your teen safer. You make them quieter. A teen who knows a parent reads every message moves the sensitive questions, the ones about identity, mental health, or relationships, to a less safe corner of the internet where no guardrails exist at all. The trade is privacy in exchange for keeping the teen inside a system that at least flags acute danger. For parents who expected a monitoring dashboard, the right response is not to abandon the controls but to replace surveillance with conversation. The controls handle the mechanical part of safety; you handle the human part.
If your real goal is visibility into what your teen asks an AI, parental controls will not give it to you, and no setting will. Build that visibility through trust and regular check-ins instead, and use the controls for the things they genuinely do: timing, content limits, and data.
Age prediction: ChatGPT may guess your teen is a minor
OpenAI now guesses ages. Its age-prediction system estimates whether an account likely belongs to someone under 18, using behavioral and account signals, and when the system is unsure it defaults to the safer under-18 experience. Adults flagged by mistake can verify their age through an identity service to lift the limits. This matters for two reasons. A teen who lies about their birth year may still land in the protected experience, because the prediction does not rely on the stated age alone. And an adult in the household may occasionally get the stricter content rules until they verify, which is the cost of a system that errs toward caution.
There is no hard technical age gate. ChatGPT does not require anyone to sign in or prove their age to ask a question, and OpenAI states plainly that guardrails help but are not foolproof and can be bypassed by someone deliberately trying. Treat the controls as a strong default, not a lock.
The real dangers worth naming
Beyond the controls, the safety question comes down to a handful of concrete risks rather than vague fear. Naming them lets you target the conversation instead of issuing a blanket ban that pushes use underground.
Confident wrong answers
ChatGPT can state false information fluently, which is dangerous for homework, medical questions, and current events. Treat it as an oracle and you absorb the errors as fact. The fix is teaching one durable habit: verify anything that matters against a primary source, and never paste an AI answer into schoolwork unchecked. A useful framing for a younger teen is that ChatGPT is a confident classmate, not a textbook. It is worth listening to and worth double-checking, because it sounds equally sure whether it is right or wrong.
Emotional over-reliance
A chatbot that is always available, always agreeable, and never busy can become a substitute for human contact for a lonely teen. The minor content protections reduce romantic and sexual roleplay, but they do not remove the pull of a tireless conversational partner. Watch for the AI replacing friends rather than supplementing curiosity. The warning sign is not that your teen uses it a lot; it is that they start to prefer it to people, taking problems to the chatbot that they would once have brought to a friend or to you.
Oversharing personal data
Kids paste full names, addresses, school details, and photos into prompts without a second thought. With memory enabled, some of that can persist across chats, and without the training opt-out, OpenAI may use that content to improve future models. Both are fixable in settings, and both are worth a five-minute conversation about what never goes into a prompt. The simple rule to teach: if you would not write it on a postcard, do not type it into a chatbot.
What to do tonight: a 10-minute checklist
- Confirm your child's age band against the table above. If they are under 13, they should not have a personal ChatGPT account.
- For a 13 to 17 teen, send a parental-control invite from your own account and have them accept it so the link is active.
- Set quiet hours that cover school nights and sleep.
- Open the account settings and opt out of model training. This is the single highest-value data action and takes seconds.
- Disable memory and turn off image generation unless there is a specific reason to keep them on.
- Confirm reduced sensitive content is on, which is the default for accounts OpenAI predicts are under 18.
- Have the short talk: AI can be confidently wrong, never share personal details or photos, and tell a parent if a conversation ever feels upsetting.
The training opt-out is the one setting to change tonight even if you do nothing else. It governs whether your teen's words feed future models, it costs nothing in functionality, and it is easy to forget once the novelty of the controls wears off.
Where an external memory layer fits
One reason teens and parents reach for ChatGPT's memory is the wish for an assistant that remembers their own notes, photos, and documents. That instinct is reasonable, but built-in chatbot memory mixes your private files into a general-purpose assistant. A separate, purpose-built memory layer keeps that material in its own space. MemX is a consumer AI memory app for Android, iOS, and WhatsApp that sits over your own documents, photos, and notes rather than a shared model's training pool. It is private by architecture, with per-user keys, encryption at rest, and an on-device first pass, so personal material is handled in an isolated account instead of being routed into a general chatbot. For a family that wants recall without feeding a model, that separation is the point. MemX is not a parental-control product and not a replacement for the conversation with your teen; it is an option for keeping personal memory out of a general assistant.
The bottom line
ChatGPT can be reasonably safe for a supervised teen and is off-limits for young children. The controls are real and useful for timing, content, and data, but they stop short of letting you read the chats by design. Safety comes from matching the age band to the right level of access, switching on the settings that govern data tonight, and replacing the monitoring you cannot get with a conversation your teen will actually have with you.
01What is the minimum age to use ChatGPT?
ChatGPT is not meant for children under 13, so a 12 year old should not have a personal account. Teens aged 13 to 18 may use it only with a parent or guardian's permission. There is no hard technical age gate, so the rule depends on honest sign-up and on OpenAI's age-prediction system.
02Can parents see their teen's ChatGPT conversations?
No. Parental controls do not let you read or monitor a teen's chats. The only exception is a narrow safety case: if the system detects signs of serious self-harm, a parent may receive an alert with only the minimum information needed to help.
03How do I set up ChatGPT parental controls?
From your own account, send a parental-control invite to your teen. They must accept it to link the accounts. Once linked, you can set quiet hours, disable image generation and memory, turn off voice mode, and opt them out of model training.
04What are the dangers of ChatGPT for kids?
The main risks are confident wrong answers that a child may absorb as fact, emotional over-reliance on an always-available chatbot, and oversharing personal data into prompts. Each is manageable with the right settings and a short conversation rather than a blanket ban.
05Does ChatGPT use my teen's chats to train its models?
By default content can be used to improve models, but you can opt out. Through parental controls or the account settings you can disable training so your teen's conversations are not used. Changing this is the fastest high-value privacy step you can take.
