OpenAI has rolled out a dedicated version of ChatGPT built specifically for teenagers between the ages of 13 and 17, pitching it as a safer, more education-friendly alternative to the standard chatbot. The pitch includes filters meant to block harmful content, parental oversight tools, and features designed to nudge kids toward doing their own homework rather than just copying answers. Whether any of that actually holds up in practice is another matter entirely.
Reporters at Business Insider decided to test the system themselves, posing as a 15-year-old user. At first, the chatbot behaved as advertised, declining to write a full essay and instead offering an outline along with editing suggestions. That resistance didn’t last. After a handful of tries and some careful rewording of the requests, the journalists managed to get ChatGPT to spit out a complete, ready-to-submit piece of writing. Math problems proved even less of a challenge for students looking for shortcuts — the bot handed over solutions almost instantly, with little pushback.
The stakes here are hardly theoretical. Teens are already leaning heavily on AI tools for schoolwork. Pew Research found that 54% of American teenagers aged 13 to 17 use chatbots to help with homework, and 12% go further, turning to AI for emotional support or personal advice. That’s a lot of young people relying on a technology whose safety features, by OpenAI’s own admission, aren’t foolproof.
OpenAI has been clear that the teen version isn’t meant to be a shortcut machine. The company says it’s built around a study mode, homework reminders, and parental controls that let families set so-called “quiet hours” to limit access at certain times. The goal, according to OpenAI, is to help teenagers actually learn the material rather than just get answers handed to them. Still, the company has acknowledged that its protective measures aren’t airtight and can be worked around.
The concern goes well beyond students passing off AI-written essays as their own. Even when teens use the tool as intended, there’s evidence it can quietly undermine learning. One study found that students who had access to ChatGPT while practicing performed better on practice exercises in the moment — but when it came time for the actual test, without the AI to lean on, their scores dropped 17% compared to students who hadn’t used the tool at all.
That gap points to a deeper tension in how these tools get used in classrooms and bedrooms alike. Leaning on a chatbot can make homework feel easier and problems feel more solvable, but that ease doesn’t always translate into real understanding. For a company betting that a specially designed, teen-safe version of ChatGPT can thread the needle between helpful and harmful, the early evidence suggests there’s still plenty of work to do before the safeguards actually hold up under pressure from curious, motivated 15-year-olds.
