What is ChatGPT for Teens?

7 min read · every claim sourced below

The short version: on 18 August 2026 OpenAI split ChatGPT into two experiences. If you are under 18, or if OpenAI’s age prediction system thinks you are, your account is moved into ChatGPT for Teens automatically. You do not opt in and you cannot opt out. It adds study tools, restricts four categories of content, forbids the model from flirting with you or claiming it has feelings, and offers parents ten settings, three of which ship switched off. If it has your age wrong, the only way out is verifying with a selfie or a government ID.

What it is, in plain terms

It is not a separate app and it is not a cut-down model. OpenAI describes it as “the same capable ChatGPT” with a different set of defaults, aimed at people aged 13 to 17.

Concretely, a teen account gets:

• Study Mode, which answers with guiding questions and step-by-step support instead of a finished solution.
• Responsible homework reminders, which can spot when someone appears to be shortcutting an assignment and steer them into Study Mode.
• Quizzes and Learning Visualizations, for practising and checking understanding.
• Study Hours, a schedule where eligible new chats start in Study Mode by default. A teen can set these themselves, or a linked parent can.
• Break reminders and product cues that keep identifying ChatGPT as an AI tool.
• Sensitive-image upload reminders, shown before some image uploads.
• Teen-specific onboarding and learning-focused starter prompts.

It is rolling out globally from 18 August 2026 on Free and paid personal plans. Full availability in Australia is expected 8 September. In the EU, age prediction arrives later to account for regional requirements.

How an account ends up in it

This is the part the announcement covers in a single sentence, and it is the part most people will want to understand. There are two routes in, and neither of them involves choosing.

The first is simple: you stated an age between 13 and 17. The second is an age prediction system. OpenAI lists the signals it uses:

• the general topics you talk about
• the times of day you use ChatGPT
• how and when your account is used
• how long your account has existed

OpenAI is also upfront that the system gets things wrong: “No system is perfect.” There is no toggle to decline it. The only exit is age verification, which is further down this page.

What is restricted

OpenAI names four categories that a teen account handles with extra care and reduced exposure:

• graphic violence or gore
• viral challenges that could push risky or harmful behaviour
• sexual, romantic, or violent roleplay
• content promoting extreme beauty standards, unhealthy dieting, or body shaming

The announcement adds that product-level interventions apply in higher-risk areas including self-harm, eating disorders and dangerous activities, and that new under-18 evaluations covering these areas have been added to OpenAI’s published system cards.

The rule almost nobody has read

The genuinely new thing here is not the content filter. It is a change to how the model is allowed to relate to a child. The updated under-18 model spec goes past blocking romantic roleplay. ChatGPT:

• should not use romantic language or terms of endearment
• should not encourage emotional dependence
• should not imply it has feelings or consciousness

OpenAI frames the purpose as reinforcing real-world relationships. Customisation is deliberately kept cosmetic for the same reason: accent colours and voice variations are offered so the experience feels personal “without blurring the line between a useful tool and a human relationship”.

What parents actually get

Parental controls are optional. The protections above apply whether or not a parent ever appears. If a parent does link an account, on the web via a phone number or email address, they can manage ten settings. Here they are with the defaults OpenAI publishes:

* May be disabled by default in some regions.

Seven of the ten ship on. The three that ship off are Study mode, Study hours and Quiet hours, which are the three the launch is built around. If a parent links an account and changes nothing, none of the study scheduling or the time limits are running.

Worth knowing separately: Improve the model for everyone is on by default, so a teen’s conversations feed model training unless someone turns it off. If that matters to you, our memo on what ChatGPT actually knows about you walks through the same controls on an adult account.

What parents never get

A parent can receive a safety notification in limited situations: when specially trained reviewers identify a serious self-harm concern, or when the teen’s account is deactivated over conversations relating to acts of violence. OpenAI is adding notifications related to eating disorders too.

A notification contains:

• a brief description of the type of safety concern
• information about an account deactivation, where relevant
• support resources or how to contact OpenAI Support

It does not contain the conversation, the transcript, or the chat history. OpenAI Support cannot supply those either. Safety notifications are explicitly not real-time monitoring.

Two more things worth knowing. The teen can start the link themselves, using an “Invite a parent” button in their own settings. And either side can unlink at any time; if the teen unlinks, the parent is notified, and after that the parent can no longer manage settings or receive safety notifications.

If it has your age wrong

You do not need to verify your age to keep using ChatGPT with teen protections in place. You only need to verify if you are 18 or over and want them removed. Here is the whole route:

On the privacy of that exchange, OpenAI is specific. Persona deletes your uploaded ID or selfie within 7 days. OpenAI never receives the document, only your date of birth or an age prediction. You may cover ID numbers such as a national ID or Social Security number before uploading, since only your name, date of birth and photo need to be visible.

There is a useful side effect: once you have verified, OpenAI stops running age prediction on your account entirely.

Three details that are easy to miss

No ads. The age prediction page states that OpenAI will not show ads in an account placed in the teen experience. It is one line in a support article, published the same week ChatGPT Ads expanded across Europe.

Turning 18 dissolves the link. When OpenAI identifies an account holder as 18 or over, the account leaves the teen experience, parental controls are removed, the connection ends and safety notifications stop. Both people are notified. A parent can invite again, but the adult has to accept.

The 90-minute break reminder is not in the docs. Several outlets reported that teens get a break reminder every 90 minutes. That figure appears in none of OpenAI’s own pages, which say only that reminders “may encourage breaks”. We have left the number out rather than repeat it.

The other half of this story

Consumer AI is being split by age. School AI is being handed out by the district. In the same month OpenAI moved under-18s into a separate ChatGPT, Anthropic made the paid Claude free for entire US K-12 schools — with an administrator, not a teacher, signing the student data agreement. If your school is anywhere near that decision, it is worth reading what Claude for Teachers gives a whole school, and the deadline on the free year.