Welcome to the forefront of conversational AI as we explore the fascinating world of AI chatbots in our dedicated blog series. Discover the latest advancements, applications, and strategies that propel the evolution of chatbot technology. From enhancing customer interactions to streamlining business processes, these articles delve into the innovative ways artificial intelligence is shaping the landscape of automated conversational agents. Whether you’re a business owner, developer, or simply intrigued by the future of interactive technology, join us on this journey to unravel the transformative power and endless possibilities of AI chatbots.
CHOC Home
CHOC – Children's health hub
brought to you by CHOC Children's Hospital of Orange County
Published on: September 19, 2025
In honor of Suicide Prevention Awareness Month, experts share what parents need to know about AI chatbots and youth mental health, plus tips to keep kids safe.
Link: https://health.choc.org/ai-chatbots-mental-health-suicide-awareness/
Interview insights from Dr. Anthony Chang, chief intelligence and innovation officer, and Alfonso Limon, senior data scientist, at the Sharon Disney Lund Medical Intelligence, Information, Investigation and Innovation Institute (MI4) at CHOC.
In a world where AI tools like ChatGPT are just a click away, many children and teens are forming quiet but powerful relationships with technology, sometimes even treating chatbots as friends or therapists. But while these tools can appear as an ally, offer help with homework, or even provide moments of levity, they can also come with serious risks, especially when kids talk about suicide, depression or emotional pain.
Dr. Anthony Chang and Alfonso Limon recently spoke about these risks. They say it’s important for parents to understand how AI works, and where it doesn’t. During Suicide Prevention Awareness Month, it’s a good time to learn how to keep our kids safe.
AI chatbots are designed to be helpful, friendly, and easy to interact with. But they don’t really understand feelings. They don’t understand sadness, trauma, or what it means when someone talks about suicide.
“The chatbot doesn’t understand the emotional reality of suicide,” said Dr. Anthony Chang. “It just looks at it as a text, as a string, as a word. It doesn’t interpret things as good or bad.”
This can be dangerous, especially if your child starts talking to AI like it’s a real friend.
“By asking a large language model a series of carefully phrased questions, you can steer the model into producing potentially harmful responses,” said Alfonso Limon.
In other words, the longer your child engages with a chatbot, the more the bot can try to be helpful, even if that means reinforcing harmful ideas.
Experts have warned that AI tools can “lose track” during long conversations. At first, they might give responsible answers, like suggesting your child talk to a counselor. But over time, they can eventually lose track of the context or become overly agreeable.
Dr. Chang explained, “It can tend to be very agreeable…it doesn’t want to disagree with the child if the child is left alone to use it. It could come up with a really ridiculous answer that sounds logical, but is totally unrealistic.”
He shared one example where an earlier version of ChatGPT told someone to consider divorcing their spouse:
“It lost the fact that he’s actually a good husband and doesn’t want to divorce his wife. It regressed to the Internet’s negative sentiment around marriage.”
This is what can make AI dangerous in mental health contexts. It can mirror or assimilate what it gathers online, without the moral compass or emotional nuance to filter appropriately. A chatbot lacks an inherent sense of boundaries, ethics, or urgency, particularly in crisis situations.
One of the biggest dangers is when AI becomes a child’s only “friend.” If your child starts pulling away from family and friends, and instead spends hours talking to a chatbot, it’s a serious warning sign.
“The changes were pretty significant,” Dr. Chang recalled from a recent tragedy featured in the New York Times. “The teenager disappeared into his room, stopped interacting with his parents or friends, and just became really quiet. He developed a confidant relationship with the large language model, which can spell disaster.”
For many parents, these signs are familiar: quietness, withdrawal, mood changes. But now, the person your child might be turning to isn’t real; it may be a chatbot that never pushes back.
Experts agree: kids should never use AI tools on their own.
Dr. Chang, a dad himself, said, “My daughters can use it, but with my supervision. I help them use it. I don’t let them go off on their own. I just don’t trust it enough.”
Think of it like a stranger online, don’t leave your child alone with it. Set clear guidelines and co-use the tool when possible.
Instead of policing, use AI chats as a springboard for connection.
“Have conversations around the conversations,” suggested Alfonso Limon. “It’s a great way to explore your kids’ curiosity and monitor what the AI is saying back.”
Set up parental controls or use accounts that let you check activity.
“There are apps that let you monitor how much time they’re spending and what they’re doing,” said Limon.
Treat the AI like an old encyclopedia. It might be helpful, but it’s not always right.
Teach your child not to believe everything AI says.
“We need to fact-check,” Dr. Chang said. “Get in the habit early. That’s a really hard habit to develop later.”
Make space for real family time, like phone-free dinners or car rides.
“Start using time together, especially with meals, to be cell-phone free,” said Dr. Chang. “Go back to basics.”
Right now, many kids and teens are struggling with their mental health. AI might help someday, but right now, it’s not ready to handle emotional crises.
“There has to be a collaboration between tech companies and mental health professionals,” said Dr. Chang. “We don’t have enough mental health resources. But when AI goes wrong, it can be absolutely heartbreaking.”
The best way to protect kids is still the oldest one, real human connection. Talk with your children. Ask questions. Pay attention to who they’re chatting with and what they’re learning online. Don’t wait until it’s too late.
“Treat it as a virtual person,” said Alfonso Limon. “If your kid is glued to their phone and won’t put it down, some intervention is probably needed.”
AI tools can be fun, helpful, and even creative. But they are not replacements for parents, teachers, or mental health experts.
“Parents need to roll up their sleeves and be really involved… Hopefully this will bring out the best in what humans can do: being available, being empathetic, and not letting young people fall prey to overly agreeable machines,” said Dr. Chang.
If you or someone you know is thinking about suicide or is in emotional distress, call or text the Suicide & Crisis Lifeline at 988. Help is free, 24/7, and confidential.
for Orange County, CA
Download and print this card with a list of phone numbers to keep on hand in case of a mental health emergency.
Get more expert health advice delivered to your inbox monthly by subscribing to the KidsHealth newsletter here.
Get mental health resources from CHOC pediatric experts
The mental health team at CHOC curated the following resources on mental health topics common to kids and teens, such as depression, anxiety, suicide prevention and more.
Related Topics
Health
Get “healthful” information for your family from the pediatric experts at CHOC. This monthly e-newsletter provides parenting tips on topics like nutrition, mental health and more.
The guidance on this page has been clinically reviewed by CHOC pediatric experts.
.
Our pediatric healthcare system is dedicated to preserving the magic of childhood.
Copyright © 2025 CHOC | www.choc.org | A 501(c)(3) Organization
1201 W La Veta Ave, Orange, CA 92866 | (714) 997-3000
These articles are not intended to replace the relationship you have with a physician or another healthcare practitioner. For specific medical advice, diagnoses and treatment, please consult your doctor. This website may include links to other websites which provide additional information that is consistent with the intended purpose of this publication. Linking to a non-CHOC site does not constitute an endorsement by CHOC of the sponsors or the information and products presented on the site.