ChatGPT for Teens is rolling out. Try one homework prompt together.

The teen experience is meant for ages 13 through 17. OpenAI says it will apply automatically when someone gives an under-18 age at signup or when its age-prediction system estimates that the account belongs to a minor. The global rollout is expected to take about two weeks, so two teenagers in the same house may not see it on the same day.

The safety changes cover more than blocked words. OpenAI says the teen model should not call itself a friend, claim personal feelings or imply that it is conscious. It also adds stronger restrictions around self-harm, graphic violence, eating disorders, risky viral challenges and romantic or sexual conversations. Teens will see more frequent reminders that ChatGPT is AI and that they do not need to keep answering it.

The learning side includes Study Mode, which uses guiding questions and step-by-step help instead of jumping straight to a finished answer. Study Hours can make that mode the default during chosen times. Quiet Hours can make ChatGPT unavailable altogether. Quizzes, visual explanations and homework reminders are also part of the teen experience.

Parents with linked accounts can manage some settings and receive limited safety notifications in serious situations. Linking is not automatic: the teen must accept the connection, and OpenAI says the parent is notified if the accounts are disconnected. These controls are a backstop, not a live window into every conversation.

A parent needs to know which features are on for this account, not which features appeared in a launch post. Study Mode does not become the default all day just because the account is marked as a teen account. A teen or linked parent has to set Study Hours if they want that behavior during homework time.

Memory deserves the same plain check. OpenAI's parental controls let a linked parent turn memory off, but the company does not list memory-off as a universal teen default. If memory stays on, ask the teen what they would be comfortable having the chatbot carry into later conversations. A useful study preference is different from a private family problem or a health worry.

Voice mode, image generation and use of conversations for model training can also be managed through parental controls. There is no single correct bundle for every teenager. A sixteen-year-old using image generation for an art class and a thirteen-year-old talking to a human-sounding voice late at night are not the same situation.

Age prediction is another reason to inspect the account instead of trusting the rollout. OpenAI says it will take the safer route when it thinks someone is under 18. Reporting from Engadget and TechCrunch notes the other side of that problem: teenagers are good at getting around age gates, and the new system has not yet had much public testing.

Start with the account. Open ChatGPT together and check whether the teen experience is active. If the rollout has not reached the account yet, say that plainly and come back to it. Do not assume the protections are present because the news says they launched today.

Then use a real school task. Ask for help with one question the student has already attempted. Try, ‘Here is where I got stuck. Give me one hint, then ask me what I would do next.’ After that, ask for the finished answer. The point is not to trick the chatbot. It is to let both of you see how Study Mode behaves when the easy shortcut is one sentence away.

Open the family controls next. Decide together whether memory, voice and image generation fit how the account is being used. Set Study Hours around actual homework time rather than making the whole evening feel monitored. Put Quiet Hours where sleep or family time keeps getting eaten. A schedule the teenager helped choose is more likely to survive the second week.

Finally, name the handoff. ChatGPT can explain a math step or turn notes into practice questions. It should not become the only place a teenager takes a crisis, a frightening health concern or a problem with someone at home or school. Pick the real people they can contact before a hard night: parent, relative, teacher, counselor, coach or another trusted adult. An app alert is late by definition; it starts after the system thinks something serious may already be happening.

Jun Vega would skip the feature tour and open the assignment already causing trouble. Ask for a hint, ask for the answer and watch what changes when Study Mode is active. If the family cannot tell whether the mode is on from the screen where the work happens, a settings page will not make the boundary feel real.

Mara Vale would not describe a parent notification as prevention. It depends on the account being linked, the system detecting the right signal and a human review happening after the conversation has already crossed a serious line. That can still help. It should sit behind a teenager who already knows which real person to contact.

Jun is testing the ordinary Tuesday-night use. Mara is testing the worst night. A family needs both: an AI tool that helps without doing the whole assignment, and a clear route away from the tool when the problem is no longer homework.

OpenAI says safety work will keep changing. Families should expect the product to change too. Recheck the settings after major app updates, after a birthday that moves the account out of teen protections, or when the teenager starts using ChatGPT differently. A setup chosen for ninth-grade homework may be wrong for college applications, a first job or a difficult personal stretch.

The company has also made big claims about learning use. Its July post says nearly nine in ten teens on ChatGPT use it for learning, information, skills or productivity in a week, and that early Study Mode evaluations showed positive results. OpenAI did not publish the trial numbers in the Engadget interview. Treat Study Mode as something to test with schoolwork, not proof that more chatbot time automatically means more learning.

The practical standard is boring. Can the teenager explain the problem after using it? Did the tool help them make a study guide, find the missing step or practice the part they did not understand? Can they close it when homework is done? If yes, the AI helped. If every assignment ends with a polished answer nobody can explain, no parental-control screen fixes that.

ChatGPT for Teens gives families a better place to start than the adult product did. Start there together. One real prompt will tell you more than a page of reassuring switches.