ChatGPT for Teens: What Changes When It Decides Your Kid Is Under 18

What this actually is

OpenAI built a system that predicts whether an account belongs to someone under 18. If it predicts yes, the account moves into what OpenAI calls ChatGPT for Teens. The account stays active, nothing gets locked, and there are no ads.

It runs even if our kid already gave a date of birth. OpenAI calls it an extra safety check, which is another way of saying the sign-up form never really worked. It is the first time a platform our kids use has said that out loud.

What it looks at

OpenAI lists the signals openly: the general topics someone talks about, the times of day they use it, how and when the account gets used, and how long the account has existed.

It is reading the pattern, not the content of any one chat. A homework question at four in the afternoon, on an account made in September, looks different from a work account made three years ago. None of those signals is about our kid personally. All of them together are a guess about our kid personally.

What changes when it decides "under 18"

A setting called Reduce sensitive content turns on, and our kid can't turn it off. With it on, ChatGPT answers more carefully around:

• •Graphic violence or gore
• •Viral challenges that could encourage risky or harmful behavior
• •Sexual, romantic, or violent roleplay
• •Content that promotes extreme beauty standards, unhealthy dieting, or body shaming
• •Other sensitive or age-inappropriate content

That is OpenAI's own list, word for word. What it does not do is hand us a transcript, or tell us one thing our kid asked.

How to check what it decided

There is now a place to look. On our kid's account, open Settings, then Account, and find the line called Age. It shows whether ChatGPT is treating that account as under 18 or as an adult. Look at it together, not behind their back. It is a question, not a confrontation.

When it gets it wrong

It gets it wrong in both directions, and OpenAI says so plainly. An adult account can land in the teen experience by mistake.

If someone 18 or older wants those protections removed, or wants it to stop predicting at all, they verify their age: a live selfie or a photo of a government ID, through a company called Persona. Persona deletes it within seven days, and OpenAI never sees the ID itself. So the only way to stop a company guessing our age is to prove it. That is a real trade, and worth knowing about before somebody in the house hits that button.

The guess and the link are not the same thing

If we have already set up parental controls, this is the question we are about to have, because linking accounts also switches on teen protections. They get described in almost the same words, and they are different things.

The guess is automatic. Nobody asked us. It reads the account, applies a content filter, and it can be wrong in either direction. The link is a choice we made, and our teen accepted the invitation. It gives us quiet hours, content filters and a set of settings we control.

The guess is a filter. The link is a control panel. The guess can switch on without us. The link needs both sides to say yes. So if the guess is already running on our kid's account, that doesn't mean anything is set up. Most of the time it means nobody has. We have a floor. The setup is still ours.

The day it all expires

When ChatGPT decides the account holder is 18, teen protections come off by default. If we linked parental controls, that link ends. Not paused: ends. We stop managing the settings and stop getting safety notifications. OpenAI notifies both of us about a week ahead, and again around the day itself.

We can send another invitation afterward, but by then they are an adult who doesn't have to accept it. And if the guess decides they are 18 before they are, only our kid can fix it. A parent can't verify a teen's age on their behalf.

So the setup we do at 13 has an end date we didn't choose. That is not a reason to skip it. It is a reason to have the conversation early enough that it still holds after the controls stop.

Why it matters for us

Pew Research Center, which is non-partisan and sells nothing, found that 68% of teens aged 15 to 17 use AI chatbots, and so do 57% of 13- and 14-year-olds. 16% use them several times a day or almost constantly. This isn't a system guessing about somebody else's kid.

Source: Pew Research Center, survey of 1,458 U.S. teens aged 13 to 17, published December 2025.

Three things to do this week

1. 1
   Link the accounts

   On a computer, open ChatGPT, click your name in the bottom left, then Settings, then Parental controls, then Add family member. Enter your teen's email or phone number, and under "This person is," pick My child. Send stays gray until you do. About two minutes, and the safeguards stop depending on a guess.

2. 2
   Ask which birthday they entered at sign-up

   If it made them an adult, that is worth a calm conversation.

3. 3
   Tell your kid the guess exists

   Not as a warning, as a fact about the tool. A kid who knows the system reads patterns understands it better than most adults do.

• ChatGPT parental controls in 15 minutes — every screen of the setup, step by step
• Is ChatGPT safe for kids? — what it is, how teens use it, and what is right at each age
• Is Character.AI safe? And 6 other AI apps — the AI tools most kids already have, and what to switch off