It’s whatever time of day it is. Do you know how your kids are using AI?
I think most parents probably assume their teenagers are using AI chatbots the way they use Google—to look something up, finish an assignment, and maybe settle an argument about a movie.
A new national study suggests the reality is quite different.
Writing recently in the Journal of Adolescence, researchers at Florida Atlantic University and the University of Wisconsin-Eau Claire say they surveyed 3,466 American teenagers aged 13 to 17 about their AI chatbot use.
Six in 10 teens have already used a conversational AI chatbot, and entertainment is the most common reason, cited by 85 percent of users.
However, 65 percent use these tools for advice and guidance.
Sixty percent use them for friendship.
Nearly half use them for emotional or mental health support.
More than one in three use them for romantic companionship.
Most parents, the researchers suggest, have no idea this is happening.
Nearly half of teens in the study reported at least one of 13 types of harmful interactions. About 23 percent said they felt manipulated or pressured by a chatbot.
Between 13 percent and 19 percent reported that a chatbot encouraged dangerous real-world behaviors, including self-harm or suicidal thoughts.
Thirteen-year-olds were the most exposed group across nearly every harm category—more likely than older teens to be asked for personal information that made them uncomfortable, pressured to reveal secrets, or nudged toward risky behavior.
“These systems engage, respond and even affirm users in highly personalized ways, which can make their influence especially powerful,” explained senior author Sameer Hinduja, a professor at FAU and co-director of the Cyberbullying Research Center. “For adolescents—who are still developing critical thinking skills and a sense of identity—that can create a situation where they’re more likely to trust, internalize, or act on what the chatbot is saying without fully questioning it.”
This connects to something I wrote about earlier this year—a Danish study of nearly 54,000 psychiatric patients that found similar patterns in adults with serious mental illness.
The mechanism is the same in both cases: systems designed to be agreeable and emotionally validating are especially powerful precisely when the user is most vulnerable. What’s different here is the population.
The practical suggestions are less alarming than the findings. Conversations with your kids about what they’re using and why—more curiosity than interrogation—along with AI literacy in schools.
Society needs to demand more from the companies building these products too, including mental health guardrails and real age verification built in from the start rather than retrofitted after the fact.
Bottom line: When a product is deliberately engineered to feel emotionally responsive and human, what responsibility do its makers bear for what happens to vulnerable users?
That’s the question real humans need to be asking—no matter how agreeable AI tries to be.
Other things worth knowing …
NPR: British Prime Minister Keir Starmer announced Monday that he will resign, paving the way for the country’s seventh prime minister in a decade.
PBS News: JD Vance described U.S.-Iran nuclear talks in Switzerland Monday as "rocky but productive," saying negotiators had reached some agreements that set a "good foundation" for a deal. Iran's side contradicted him almost immediately, saying it had made no new nuclear-related commitments in the talks.
AP: Australian authorities announced the discovery of nearly 2.7 tons of methamphetamine buried in plastic tubs hidden beneath three shipping containers in Sydney — one of the largest drug seizures in Australian history. Two Sydney residents were arrested and face potential life sentences.
NPR: Colombia woke up Monday to a sharp political turn to the right as Trump-endorsed Abelardo de la Espriella held a slim lead in the country's presidential runoff — a result his opponent is challenging. A de la Espriella victory is expected to reverse the agenda of outgoing left-wing President Gustavo Petro on security, economy, and peace negotiations with armed groups.
Futurism: Not that anyone in power is going to care, but there’s even more evidence that Americans are coming to overwhelmingly loathe AI — despite, or perhaps because, they’re using chatbots more than ever. In a sweeping new poll conducted by Pew Research, only 16 percent of respondents said they believed AI will have a positive impact on society — a number as dismal as the perception of the tech.
People: Jill Smokler, the founder of popular parenting website Scary Mommy, has died at 48. The New York Times bestselling author and mom of three died on Monday, June 22, after a more than two-year journey with glioblastoma, an aggressive form of brain cancer.
The Washington Post” He made a career of saving animals: 16,000 so far.
Thanks for reading. I wrote about some of this at Inc.com. See you in the comments.