More and more teens are interacting with chatbots like Open AI’s ChatGPT, Character.AI and Meta AI.
A majority, 72% of teens ages 13 to 17 have used AI companions at least once, according to a July 2025 report by nonprofit Common Sense Media. More than half, 52% interact with the platforms at least a few times a month, and 13% are daily users.
While teen use of chatbots can be fairly benign — 46% say they use them as a tool or program — for some, the reliance and relationship can go much deeper, sometimes to tragic ends.
Last week, parents of teens who died by suicide after extensive chatbot use testified before Congress about the dangers of this new tech.
NYU Stern School of Business professor and bestselling author of “The Anxious Generation” Jonathan Haidt has spent the last few years sounding the alarm about teen tech use.
He’s also been giving parents actionable steps they can take to protect their kids’ mental health.
‘No chat can go longer than 30 rounds’
AI chatbots are “incredibly dangerous,” Haidt told CNBC Make It at the Fast Company Innovation Festival last week.
“We have deaths. We have delusions in adults as well. And right now, the most dangerous thing seems to be the relationships, the long conversations.”
If kids are using AI as a tool to learn and to find information it’s “generally a good thing,” Haidt says. In fact, by high school, they’ll probably have to use it for assignments.
The problem arises when the use of this tech takes a wrong turn and kids start to build relationships with it. That’s where, Haidt says, it’s prudent for parents to set boundaries.
To ensure kids are only turning to AI chatbots as a tool, parents can set rules about how they use them at home. They can restrict use only to a shared device like a family computer.
No children should be having a relationship with AI.
Jonathan Haidt
Professor, bestselling author
Haidt advises parents set a limit on how long their kids can engage in “conversation” with a chatbot and suggests no more than 30 rounds of back and forth. In the stories that ended tragically, conversations were “1,000s and 1,000s of rounds” and that’s what made the difference, he says.
In an August 2025 blog post that addressed the rapid adoption of ChatGPT and the app’s safety, OpenAI wrote “our safeguards work more reliably in common, short exchanges. We have learned over time that these safeguards can sometimes be less reliable in long interactions.”
Haidt contends that because tech companies “have a long track record of harming children at an industrial scale,” the onus is on parents to enforce strict rules around using their products.
“No children should be having a relationship with AI,” he says.
Want to be your own boss? Sign up for Smarter by CNBC Make It’s new online course, How To Start A Business: For First-Time Founders. Find step-by-step guidance for launching your first business, from testing your idea to growing your revenue. Sign up today with coupon code EARLYBIRD for an introductory discount of 30% off the regular course price of $127 (plus tax). Offer valid September 16 through September 30, 2025.
Plus, sign up for CNBC Make It’s newsletter to get tips and tricks for success at work, with money and in life, and request to join our exclusive community on LinkedIn to connect with experts and peers.