ChatGPT for Kids: Is It Safe and How to Set It Up

There is no ChatGPT for kids. OpenAI does not sell a child account, a simplified interface or an under-13 tier, and its terms set 13 as the hard minimum age with parental permission required for anyone under 18. What does exist, since 18 August 2026, is ChatGPT for Teens, a named experience for 13 to 17 year olds that applies itself automatically, plus a parental controls panel that has quietly grown from five settings at launch to nine today.

Most guides still describe the September 2025 version of this feature. That means they are missing four of the nine controls, the age prediction system that went live in January 2026, the two updates OpenAI shipped in July 2026, and the ChatGPT for Teens launch on 18 August 2026. This guide covers what parental controls actually do, the exact steps to switch them on, and every setting with its default state. It also covers the one thing parents keep assuming they can do and cannot, plus the workarounds your teenager will find within a week.

The Key Takeaways

• ChatGPT for Teens launched on 18 August 2026. There is nothing to download and nothing to buy. If OpenAI estimates an account holder is under 18, or they state an age of 13 to 17, it becomes their default experience.
• 13 is the floor. Under-13s are not permitted on ChatGPT at all, and no supervised option exists for them the way it does on Google Gemini.
• Parents cannot read the conversations. There is no transcript view, no activity log and no weekly summary email. OpenAI states this twice in its own documentation.
• There are nine parent-managed settings, not the five most articles list. The newest cover Study Mode, saved memories, Codex network access and the cloud browser.
• Age prediction has been live since January 2026 and can override the birthday your teen typed in. Getting out of the teen experience requires an ID check through Persona.
• Controls are opt-in on both sides. Your teen has to accept the invite, can unlink at any time, and loses all of it automatically at 18.

Is ChatGPT Safe for Kids?

ChatGPT is not designed for children. OpenAI sets 13 as the minimum age and requires parental consent under 18. For 13 to 17 year olds it is reasonably safe once a parent links accounts, which switches on content safeguards automatically. For under-13s there is no supervised option at all, so a different tool is the safer choice.

That is the honest version. The longer answer depends on one thing, which is whether the account is linked.

An unlinked teen account and a linked one are meaningfully different products. Linking triggers a safeguard called Reduce sensitive content that is on by default and cannot be switched off by the teen.

That safeguard targets four specific categories, and OpenAI names them plainly. It reduces graphic content, viral challenges that push risky behaviour, sexual, romantic or violent roleplay, and content promoting extreme beauty ideals or unhealthy dieting. Those are the risks OpenAI decided were worth engineering against, which tells you something about what teenagers were actually running into.

The gap that remains is emotional rather than technical. A chatbot that responds instantly, never judges and is available at 2am is a different kind of risk from a website with bad content on it. OpenAI’s advice on its parent resources page, echoed by Common Sense Media, is that controls work best alongside an ongoing conversation, not instead of one. The controls arrived after real harm, including the teen suicide case that pushed OpenAI to act, and OpenAI built them in consultation with the Attorneys General of California and Delaware.

There Is No ChatGPT for Kids, and That Matters

Searches for a kids version of ChatGPT turn up a lot of confident answers describing a product that does not exist. There is still no separate app for children, no child mode, no simplified interface and no free kids tier. OpenAI has never shipped one.

What changed on 18 August 2026 is that the teen experience stopped being an unnamed set of defaults and became an announced product called ChatGPT for Teens. That is the thing a lot of those pages were vaguely gesturing at, and it is worth being precise about who it covers. It serves 13 to 17 year olds, which is the age band that was already permitted on ChatGPT. It does not lower the age floor, and it does nothing for a nine year old.

Be especially careful with pages describing an “OpenAI Family Plan” with content filter levels labelled strict, moderate and standard, or a weekly activity summary emailed to parents. None of that appears anywhere in OpenAI’s documentation, and the weekly summary directly contradicts OpenAI’s stated position that parents do not get visibility into conversations. Treat those pages as unreliable on everything else too.

So the whole system is two things: ChatGPT for Teens, applied automatically to accounts identified as under 18, and parental controls layered on top once accounts are linked. If your child is under 13, the correct answer is not a workaround, it is a different tool.

What ChatGPT for Teens Actually Changed

OpenAI began rolling ChatGPT for Teens out globally on 18 August 2026 and said the rollout would finish over roughly the following two weeks. There is no new app and no new account. Lauren Jonas, OpenAI’s head of youth and families, put it plainly to Engadget: if the system predicts you are under 18, or you have said so, this simply becomes your default experience.

Most of the parts were already shipping separately over the past year. What is new is that they are bundled, on by default, and joined by a handful of genuinely new features.

The Learning Half

The headline feature is Study Hours, which lets a teen or their linked parent choose times of day when Study Mode switches on by default. It turns a per-chat toggle into a routine, which is the first version of this that survives contact with an actual teenager.

Alongside it sit responsible homework reminders, which fire when the model detects that a teen appears to be shortcutting an assignment and redirect them toward Study Mode. Quizzes and learning visualizations round it out, the latter now covering more than 300 topics against the 70 or so it launched with in March 2026.

One correction worth making, because a lot of the coverage got it backwards. ChatGPT for Teens does not default to Study Mode. Jonas confirmed as much. The starter prompts are learning-focused and the reminders nudge, but the mode itself is still opt-in unless Study Hours or the parental control is set. If you want it on, you still have to turn it on.

The Safety Half

Protections come on by default and follow OpenAI’s under-18 principles in its Model Spec, targeting self-harm, violence, eating disorders, dangerous activities and explicit sexual or graphic content. The more interesting change is about the relationship rather than the content. The updated spec says the model should not use romantic language, encourage emotional dependence, or imply that it has feelings or consciousness, and Jonas described the rule as the model not calling itself a friend or suggesting sentience.

That is a direct answer to the risk this guide flagged earlier, the 2am companion problem rather than the bad-content problem. Break reminders now fire more often for teens and lean on the point that ChatGPT is an AI and does not need an immediate reply. Teens also get sensitive-image upload reminders, a teen-specific onboarding walkthrough, and cosmetic customization such as accent colours and voice variations, which OpenAI describes as personalisation that stops short of blurring the line between a tool and a relationship.

What Is New for Parents

Less than you might hope, and that is deliberate. The panel is the same one described below, and the ceiling on what parents can see has not moved at all. The one real addition is that safety notifications now cover eating disorders, on top of the self-harm and violent-activity categories already in place.

Jonas gave Engadget the mechanics that OpenAI’s own announcement leaves out: alerts go by text, in-app and email, every flagged item is reviewed by a full-time OpenAI employee before a parent hears anything, and the target is to notify within an hour of the prompt.

OpenAI has also started publishing under-18 evaluations in its system cards, covering self-harm, eating disorders, violence, age-restricted goods and sexual content. That is the first time there has been a public number to argue with rather than a description of intent, which matters more than any single toggle.

ChatGPT’s Age Rules and the Age Prediction System

OpenAI’s terms require you to be at least 13, or the minimum age of digital consent in your country, whichever is higher. Anyone under 18 needs a parent or guardian’s permission. Historically that was self-reported at signup, which meant it was enforced by nothing more than a date field.

That changed in January 2026 with age prediction, now rolling out globally with the EU delayed for regional requirements. The system estimates whether an account belongs to someone under 18. The signals include the general topics discussed, the times of day the account is used, how it is used, and how long it has existed. It can override the birthday your teen typed in at signup, which is the point of it.

An account placed in the teen experience stays fully active, and OpenAI says ads are not currently shown to teens on any tier. Getting out requires proving your age through Persona, a third-party verification provider, using a live selfie, a government ID, or both depending on the country. Persona deletes the upload within 7 days, and OpenAI never receives the ID itself, only a date of birth or a pass result.

One regional detail worth knowing. In Italy, users asked to verify have 60 days to complete it before some features may be disabled. Everywhere else, verification is optional and only needed by adults who were misclassified.

How to Set Up ChatGPT Parental Controls

Both of you need a ChatGPT account before you start, and your teen has to accept the invitation. You cannot set this up unilaterally. The whole flow takes about two minutes on the web.

1. Open Settings from your profile menu in ChatGPT.
2. Select Parental controls from the side menu.
3. Select Add family member.
4. Choose Use phone or Use email, then enter your teen’s number or email address.
5. Select Send. Your teen gets a text or email with a link, and may also see a notification inside ChatGPT.
6. After they accept, select their name under Family members to review and change settings.

It works in the other direction too. A teen can go to Settings > Parental controls and choose Invite a parent, or Accept invite if you have already sent one. One parent can link to several teens, but each teen can only be linked to one parent at a time.

Every ChatGPT Parental Control Setting and Its Default

This is where most guides are out of date. The original announcement listed five controls. There are now nine in OpenAI’s parental controls FAQ, and the defaults matter as much as the list, because several of the ones parents assume are off ship as on.

Your teen cannot change any of these while the accounts are linked. Changes may not apply instantly either, since some take effect only in new conversations while existing chats keep their old behaviour.

Quiet Hours

Quiet hours block ChatGPT entirely during a window you choose, including ChatGPT Work. You set a start time and an end time under your teen’s name in the Parental controls panel. The limitation to know about is that you only get one window, so you cannot block school hours and bedtime separately.

The Study Mode Toggle

Added on 16 July 2026, this is the most misunderstood control in the panel. Turning it on means every new main conversation starts in Study Mode, which walks a student through a problem step by step instead of handing over the answer. It does not block Study Mode, it does not force it, and it does not apply to GPT or Project conversations. If you want the detail on what the mode actually does, we covered how ChatGPT Study Mode works separately.

Study Hours supersedes this for most families. Shipped with ChatGPT for Teens on 18 August 2026, it does the same job on a schedule rather than as a permanent on switch, and your teen can set it themselves. The all-or-nothing toggle is still there if you want it, but a homework window is an easier thing to agree on than a blanket default. Note that OpenAI’s parental controls FAQ still lists the nine settings above and has not folded Study Hours into that list, so treat it as a companion feature rather than a tenth control.

Safety Notifications

These are on by default and narrow by design. Systems flag prompts indicating serious self-harm concerns or conversations relating to acts of violence. A small team of trained reviewers then assesses them, and the linked parent is alerted by SMS, email or push notification with support resources attached.

The 13 July 2026 update widened this to cover teen accounts banned for violent activity, and the ChatGPT for Teens launch on 18 August 2026 added eating disorders as a category. OpenAI is explicit that it is not meant to fire on fictional writing, gaming, news discussion, political discussion, general anger or abstract questions. No system is perfect, and OpenAI says so, but the design intent is a rare, serious alert rather than a monitoring feed.

What Parents Cannot Do

You cannot read your teen’s ChatGPT conversations. Parental controls give you no transcript access, no history view and no real-time monitoring. OpenAI states this twice in its own FAQ, and again on its parent resources page. When a safety notification does fire, it shares only the information needed to support the teen’s safety.

Several other limits are worth knowing before you rely on this. Your teen can unlink the accounts at any time, though you are notified when they do, and you cannot set more than one quiet hours window.

There is also an expiry date on all of it. When the account is identified as 18 or older, parental controls are removed automatically, notifications stop, and you both get told.

Adults get a separate feature called Trusted Contact, available at 18 or older, and 19 in Korea. It lets someone nominate one person to be notified if ChatGPT identifies a serious suicide-related safety concern. It is not parental controls under another name, and it gives the contact no settings access.

How Kids Get Around ChatGPT Parental Controls

The controls only govern one specific account. That is the whole weakness.

The simplest workaround is signed-out browsing. Opening chatgpt.com in a browser without logging in reaches ChatGPT with no linked controls applied at all. The second is a second account, and nothing stops a teenager creating one you have never seen. Age prediction might still catch it and apply the teen experience, but your settings will not follow it.

The third is simply unlinking, which your teen can do at any time. You get a notification when it happens, so it is visible rather than silent, but it is one tap on their side. OpenAI acknowledges the broader problem too, noting that guardrails can be bypassed by someone deliberately trying to get around them.

Treat parental controls as a default-setter for a cooperative teenager, not a lock.

Blocking ChatGPT at the Device Level

Device restrictions close the signed-out loophole, since they apply to the phone rather than the account. On iPhone, open Settings > Screen Time > Content & Privacy Restrictions > Web Content, choose Limit Adult Websites, and add chatgpt.com to the Never Allow list. Apple’s web restrictions apply across Safari and any browser that respects them, which pushes your child back into the app where your linked settings actually work. If you are setting this up on a recent device, our iOS 27 Screen Time features rundown covers where Apple moved things.

On Android, Google Family Link does the equivalent job, and on Windows or Xbox, Microsoft Family Safety can block apps and set screen time limits. Router-level or DNS blocking is the heaviest option and catches every device on the home network at once.

Is Gemini Safer Than ChatGPT for Kids?

For under-13s, yes, and it is the only meaningful difference between the major assistants. Google allows children under 13 to use Gemini Apps through a Family Link supervised account, with the parent turning access on or off and getting an email the first time the child opens it. OpenAI offers nothing comparable.

Two caveats before you switch. Gemini Apps are not available for supervised accounts in the European Economic Area, Switzerland or the United Kingdom. And Google is blunt that its filters are imperfect, warning that a child may still encounter content you would rather they did not.

Claude is the strictest of the four. Anthropic restricts its consumer product to 18 and over, asks users to verify when it detects signals of a minor, and disables accounts on indicators of underage activity.

There is no teen tier to configure, so it is not an option for a school-age child at all.

What We Would Actually Do

If your child is under 13, keep them off ChatGPT and use a supervised Gemini account through Family Link instead, assuming you are not in the EEA, Switzerland or the UK. If they are 13 to 17, link the accounts. Linking is the single highest-value action here because it switches on the content safeguards automatically, and nothing else you configure matters as much.

After linking, the three settings worth thinking about are quiet hours for bedtime, image generation if you want it off, and Study Mode if homework is the main use. If homework is the whole reason ChatGPT is in the house, set Study Hours for the after-school window instead of forcing Study Mode on permanently, and let your teen have a say in where the window sits. Leave Improve the model for everyone off, which is where it already sits. Then have the conversation, because the controls are a floor rather than a solution, and your teen can unlink them tomorrow. If ChatGPT is mostly a homework tool in your house, it is worth comparing it against the best AI tools for students, and checking what the free ChatGPT plan includes before anyone pays for anything.

FAQ