ChatGPT for Teens rolled out globally starting 18 August 2026 (Australia follows by 8 September). It isn't a separate app or a signup you choose: OpenAI's age prediction system, built from account age, usage patterns, and stated age, automatically routes anyone it believes is under 18 into a restricted experience. "If there is doubt, we'll play it safe and default to the under-18 experience," per OpenAI's own framing. The teen mode blocks romantic or sexual roleplay, restricts self-harm and graphic content, and redirects homework requests into Study Mode's hints-and-questions format instead of handing over a finished essay.
Parents link an account via Settings, but they see safety notifications and can toggle settings, not chat transcripts. If you're wrongly flagged as a teen, the fix is self-verification with ID at chatgpt.com/verify_age, not a parent's say-so.
The launch, detailed in OpenAI's own announcement, lands after months of lawsuits alleging ChatGPT contributed to teen self-harm and public pressure from state attorneys general. It's the branded, permanent version of an age-prediction system OpenAI first rolled out in January 2026, now bundled with new parental controls and homework-specific guardrails.
How the age detection actually works
There's no ID check by default. OpenAI blends three signals: your account-stated age, any age you've formally verified, and a behavioral age-prediction model that reads account age, usage-time patterns, and long-term activity. Per Lauren Jonas, OpenAI's head of youth and families: "There's no need for a teen to create a new account or change anything; if we predict you're under 18, or you've told us so, this becomes your default experience." An existing adult account gets flagged in place, not replaced.
The system errs toward caution, and that has a real cost for adults. When the underlying age-prediction model first shipped in January 2026, TechRadar reported it misfiring on new accounts (even long-time paying subscribers), late-night usage, teen-relevant topics, non-native-speaker grammar patterns getting misread as younger writing, and shared family devices. One separate figure worth knowing before you trust the system either way: a Wall Street Journal report on internal OpenAI documents, tied to the company's since-shelved "adult mode" erotica plans rather than this week's teen-mode launch, cited an estimated 12% false-negative rate, meaning roughly one in eight actual minors got classified as an adult. OpenAI hasn't published fresh misfire data tied to this week's specific launch, so treat that figure as evidence about the same underlying age-prediction model in a different product debate, not this week's news.
If you're an adult wrongly placed in teen mode, the fix is Settings → Account → Verify age, or chatgpt.com/verify_age directly. That runs an ID-plus-selfie check through a third-party vendor, Persona; OpenAI says it never sees the raw image itself, only a pass/fail result, and the data is deleted after the check. Notably, a parent cannot verify a teen's age on the teen's behalf. Only the account holder can run that check.
What's actually restricted
- No romantic or sexual roleplay, and no romantic language or terms of endearment from the model.
- The model won't claim to have feelings, consciousness, or emotions, and OpenAI says it limits behavior that encourages emotional dependence or frames itself as a friend.
- Restricted output on self-harm, suicide, eating disorders, violence, dangerous activities, and sexually explicit or graphic content.
- Homework shortcuts get redirected. A direct request for a finished essay or answer set is meant to trigger a "responsible homework reminder" that pushes into Study Mode: hints, step-by-step guidance, follow-up questions, and knowledge checks instead of the finished product.
- Break nudges after 90 minutes of activity within a 3-hour window.
Neither OpenAI nor the outlets covering the launch published an actual before-and-after transcript showing the homework redirect in action; every description of it is OpenAI's own framing of the mechanism, not an observed example. If a teen in your household hits this, expect Study Mode's format specifically (worked hints and questions), not a flat refusal.
Parental controls: what parents see, and what they don't
Linking works through Settings → Parental controls → Add family member, sent by phone or email invite, and the teen has to accept it. One parent can link to multiple teens, but each teen links to only one parent account at a time.
Parents cannot read chat content or transcripts. They can toggle settings (a reduce-sensitive-content switch, memory and data-use controls, Voice Mode, image generation, cloud browser access, Study Mode, and Quiet Hours), set Study Hours (windows where Study Mode is on by default) and Quiet Hours (blackout windows a teen can't override mid-chat), and they receive safety notifications for high-risk situations, such as self-harm content flagged by a human reviewer or an account action taken over violent content. Those notifications describe the type of concern only. They never include the transcript itself.
At 18, the teen protections turn off by default and the parental link auto-disconnects; both sides can choose to reconnect it voluntarily. What happens to chat history when someone moves into or out of teen mode isn't documented anywhere OpenAI or the outlets covering the launch addressed, including TechCrunch, which flagged the same gap. Don't assume it's preserved or wiped either way until OpenAI says so.
How this compares to Gemini and Claude
The three major assistants take genuinely different approaches, and it's worth knowing where each one sits if a household or classroom uses more than one. Anthropic's Claude doesn't have a teen tier at all: its terms require users to be 18 or the local age of majority, full stop, with no parental-consent flow, enforced through account signals plus app-store age-verification laws now required in some US states. Google's Gemini offers a separate Gemini Teen Experience tied to Family Link for under-18 accounts, but its safeguards are content-filtering and fact-checking oriented (no image generation, Search-backed grounding, safety alerts pointing to help resources) rather than built around a pedagogy layer the way ChatGPT for Teens' Study Mode redirect is. Google is separately rolling out its own automatic age-detection across Search, Play, and YouTube, region by region.
For a student doing research specifically, that pedagogy-vs-filtering split matters. ChatGPT for Teens is trying to stop a student from getting a finished essay handed to them. It isn't built to show its work the way a source-grounded tool does. If the goal is homework a teacher can trust wasn't just generated wholesale, a citation-backed research workflow, gathering the actual sources in Gemini Notebook and asking for a summary with inline citations back to the original text, gives a student and a teacher something to verify, rather than a black-box answer with no paper trail at all.
People also ask
Do teens need to sign up for a new ChatGPT account to get teen mode?
No. Existing accounts are automatically flagged in place based on OpenAI's age-prediction system. There's no separate app or signup flow.
Can parents read their teen's ChatGPT conversations?
No. Parental controls let a linked parent manage settings (content restrictions, Study Hours, Quiet Hours, memory, image generation) and receive high-risk safety notifications, but they cannot view chat transcripts or message content.
I'm an adult and ChatGPT thinks I'm a teen. How do I fix it?
Go to Settings → Account → Verify age, or visit chatgpt.com/verify_age directly, and complete an ID-plus-selfie check through OpenAI's verification vendor, Persona. A parent or anyone else cannot verify your age for you.
Does ChatGPT for Teens block homework help entirely?
No. It redirects direct requests for a finished answer into Study Mode's format, hints, step-by-step guidance, and follow-up questions, rather than refusing to help with the subject at all.
Does Claude or Gemini have something similar?
Claude has no teen tier and requires all users to be 18 or older. Gemini offers a separate Teen Experience through Family Link with content-filtering safeguards, but it isn't built around the same homework-redirect mechanic as ChatGPT for Teens' Study Mode.
When is ChatGPT for Teens available everywhere?
The global rollout started 18 August 2026 and is expected to complete within about two weeks, with one confirmed exception: full availability in Australia is set for 8 September 2026.