SAN FRANCISCO, August 18, 2026, 07:48 PDT
OpenAI launched ChatGPT for Teens on Tuesday for users aged 13 to 17. The service automatically applies tighter safeguards when an account declares, or appears, to belong to a minor.
The key change is automatic routing, not parental surveillance. Teen protections can apply without a linked adult account. Parental controls remain optional and require an accepted invitation.
That split matters for families and schools. OpenAI defaults uncertain accounts to the safer experience, while giving misclassified adults a verification route.
| Account signal | Default treatment | User action |
|---|---|---|
| Under 13 | ChatGPT is not intended for this age | An adult must conduct any education use |
| Age 13–17 declared | Teen experience | Parental consent is required |
| System predicts under 18 | Teen experience may apply | No parent link is needed |
| Age is uncertain | Safer teen setting | Adult users may verify age |
| Verified 18 or older | Adult settings | Teen protections are removed |
ChatGPT remains unavailable to children under 13. Users aged 13 to 18 need parental consent, subject to local minimum-age rules. OpenAI also warns that safeguards cannot make every answer age-appropriate.
The teen experience restricts graphic violence, self-harm depictions, risky challenges and sexual or romantic roleplay. It also responds more cautiously to eating disorders and extreme body-image content.
| Area | Teen experience | Standard adult experience |
|---|---|---|
| Sensitive content | Extra reduction is on by default | User-selectable where available |
| Romantic or sexual roleplay | Restricted | Standard policy applies |
| Emotional framing | Cannot suggest personal feelings or consciousness | Standard policy applies |
| Homework | Responsible reminders and Study Mode prompts | Study Mode remains available |
| Break reminders | More frequent | Standard frequency |
| Advertising | No ads | Not specified on the teen-policy page |
“We want to treat teens like teens,” Ann O’Leary, OpenAI’s vice president of global policy, told the Associated Press. The company also bars the bot from presenting itself as conscious or emotionally dependent on the user. Associated Press
OpenAI paired those limits with learning tools. Homework reminders steer shortcuts toward Study Mode, while quizzes and visual explanations support guided work. Study hours can activate that mode on a schedule.
“We believe AI should expand educational opportunity,” OpenAI said. It designed the service for moments outside classrooms, when help may be unavailable. Reuters
| Linked-parent control | What it changes | Important limit |
|---|---|---|
| Quiet Hours | Blocks ChatGPT during one daily window | Only one window can be set |
| Study Mode | Starts new chats in guided-learning mode | Does not guarantee academic integrity |
| Voice | Allows or disables voice use | Setting changes may lag |
| Memory | Allows or disables saved memory | Deletion may take up to 30 days |
| Image generation | Allows or disables image creation | Other age limits still apply |
| Model training | Controls use of chats to improve models | Separate from conversation access |
| Safety notifications | Alerts on reviewed high-risk signals | Not continuous monitoring |
Parents cannot read a teen’s chats through the controls. OpenAI may send limited safety information after trained reviewers assess serious self-harm or violence signals. Either side can unlink the accounts.
Age prediction uses signals such as account age, general topics and usage times. It can override a previously supplied birth date. The system is rolling out globally, with the European Union due in coming weeks.
| Verification element | Published treatment | Timing or scope |
|---|---|---|
| Age-prediction inputs | Topics, usage timing, account use and account age | Global rollout |
| Adult challenge | Live selfie and/or government ID through Persona | Available after misclassification |
| Data sent to OpenAI | Date of birth or verification outcome | OpenAI says it does not receive the ID |
| Persona retention | Uploaded ID or selfie deleted | Within seven days |
| European Union | Age prediction planned | In coming weeks |
| Italy | Verification required when requested | Within 60 days |
Risks: Age prediction can miss teens or restrict adults. Content filters can also fail. Safety alerts are limited, parent linking is optional, and control changes may not apply immediately.
The launch therefore raises the product’s safety floor without making ChatGPT a monitored service. Its practical test will be whether automatic classification improves protection without imposing excessive verification on adults.