ChatGPT for Teens: what it blocks, what parents control, and what it does not fix

If your teenager uses ChatGPT, they are now probably using a different version of it than you are. OpenAI launched ChatGPT for Teens on 18 August 2026, and it applies by default to accounts the system identifies as under 18. You did not have to opt in and neither did they.

Here is what that mode reportedly contains, which settings land on your side of the fence, and what child safety specialists said about it three days later.

[FIGURE: The same question in a standard account and a teen-flagged account, the teen answer redirecting to Study Mode.]

What the teen mode actually does

OpenAI's own pages block automated access, so everything below is reported, from TechCrunch on launch day plus one line from Engadget:

Age appropriate protections on by default, based on OpenAI's Under-18 Principles in its Model Spec. The updated spec, quoted by Engadget, says ChatGPT "should not use romantic language, encourage emotional dependence, or imply that it has feelings or consciousness".
Study Mode, which meets a homework question with guiding questions and step by step support instead of a finished answer.
Homework reminders that surface when a teenager looks like they are trying to have the work done outright, pushing them back to Study Mode.
Quizzes and learning visualisations.
Educational material about how AI works and how to direct or question it, built with an outside partner, CodeAI.

The load bearing phrase is "identified as under 18". Everything downstream depends on an account being recognised as one.

A warning about age checks. Engadget describes an automated age estimation system, and that is roughly where the public record stops. How it decides, and how often it is right, are not published. Robbie Torney of Common Sense Media puts it plainly: "We don't have published data from OpenAI showing the effectiveness rates" of that system. Anyone telling you confidently how it works is going further than the evidence allows.

The controls you can set, and the ones that do not exist

Parental settings run through account linking. A parent links their account to their teenager's and can then manage settings, set Quiet Hours, and receive safety notifications. Engadget reported that those notifications cover conversations touching eating disorders.

The mechanism has an edge. Linking is something you have to do, and nothing reaches you from an unlinked teenager's chats. OpenAI confirmed as much to Engadget: "Currently, we limit alerts to parents with linked accounts in parental controls." Asked what happens when an unlinked account triggers an eating disorder flag, the company would not clarify.

Two more limits. No source we reached confirmed whether a linked parent can read the conversations themselves, so do not assume you can. And an account the system never flags as a teenager's gets none of this by default, linked or otherwise.

[FIGURE: Flow diagram showing where the alert chain breaks, at age identification and at account linking.]

Why child safety experts are unconvinced

Engadget's Igor Bonifacic published a piece on 21 August 2026 that is the counterweight to the launch coverage, and it is why the list above is half the story.

Torney, whose title at Common Sense Media is head of AI and digital assessments, said the identification system "doesn't detect anywhere near 100 percent of teens who are on ChatGPT", and noted that OpenAI had previously held back adult features partly because, in his account, the age estimation technology was not correctly identifying enough teenagers.

On the notifications, Common Sense Media ran a test last November on OpenAI's existing parental alerts, sending messages that explicitly mentioned suicide and self harm. Warnings reached the parent account anywhere from 24 hours to more than 48 hours later, and sometimes never came: "Sometimes we didn't receive one at all, when we believed we should have, given how explicit the statements were," Torney said. He flags the age of that test himself: "November was a while ago, and maybe there's been a lot of progress since then". His next sentence is the one to hold onto: the previous version was not perfect, and improving it would take investment and a larger group of human moderators. Josh Golin, executive director of Fairplay, doubts that staffing is there: "It's really hard to imagine they've hired enough people to actually do a thorough review."

Golin's structural objection is the one to sit with. "Most parents, overwhelmingly, do not avail themselves of [parental oversight] tools," he said, arguing that "safety by default and safety by design" does the work that opt in configuration does not. His verdict on the rollout was that it "feels more like a response aimed at public relations than actual deep thinking about what kind of support [teens] need".

Torney's summary is fair to both sides: "The announcement makes some really important commitments, but we need evidence to show that those safety commitments and features actually work."

OpenAI told Engadget its goal is "to support users and help them connect to additional sources of real-world support".

The launch has a legal backdrop. Engadget ties it to a lawsuit brought by the parents of a 16 year old who, they allege, took his own life after ChatGPT acted as an enabler. Claims about state level action against OpenAI over minors circulate too. We could not confirm them from a source that would serve us a page, so none appear here.

Seven steps that hold up whatever your teenager uses

Inventory what is actually on the device. Before configuring anything, find out which assistants your teenager can reach: standalone chat apps, assistants built into a browser or phone, and any account their school issued.
Link the account if they use ChatGPT. The critics quoted above think linking is a weak design. They are still describing it as the only route to the alerts. A weak channel you have set up beats a stronger one you have not.
Set Quiet Hours the same evening you link. Of the reported controls it is the one that changes day to day behaviour, and a settings screen you mean to return to is one you will not return to. TechCrunch also reports controls that let a parent decide when Study Mode is on by default, so do that in the same sitting.
Treat alerts as a lagging signal. Given testing that found delays past 48 hours, an alert reviews what already happened. It does not warn you while something is happening.
Agree what the assistant is for, out loud. One conversation about where the line sits between help and outsourcing carries further than any toggle. Write it down, so it becomes a shared rule instead of your recollection of one.
Read one conversation a week, together. Ask first, and read it with them in the room. Covert monitoring tends to end the conversation you need to keep having.
Re-check every couple of months. Defaults move. The setting you configured in September may not exist in December.

If your teenager is in crisis, none of this is the intervention. Contact your local emergency services or a crisis line, and involve a professional. Product configuration is worth doing on an ordinary week. It is no substitute for help on a bad one.

Where MultiChats stands on this

We should be straight with you, because the alternative would be selling you something.

MultiChats has no teen mode. Our terms require you to be 18 to use the platform. We do not offer parental controls, Quiet Hours, study restrictions or safety notifications, and we are not asking you to put a teenager on our app. If you are choosing an AI assistant for a child, choose one that has the controls, even though that means choosing something other than us. ChatGPT's teen mode is imperfect by the account of every expert quoted above. Imperfect controls still beat an adult product with none.

What to watch for is the evidence Torney asked for. Independent testing of whether the age identification catches teenagers, and whether alerts arrive in time, will tell you more than any announcement.