The use of AI chatbots has skyrocketed in recent years, including by teens and young adults. It may seem harmless at first, but investigating into the habits it builds, feelings it imitates, and data it collects, it may not be as simple as it seems.
With AI’s rise in popularity, all types of people are using it for many purposes. One of the popular uses for young people is mental health support. With the adaptability of chatbots, many people turn to AI with problems they don’t feel comfortable sharing with loved ones because it can feel very similar to talking with a person without the risk of burdening or worrying the people closest to them. However, it might not be safe for young people to discuss these sensitive and personal topics with AI.
More and more teens are using chatbots. Some use them as a tool, others talk to them like a friend, but according to a study by Common Sense Media, 52% of teens are regular users of AI companions.
AI is trained to affirm the user and doesn’t understand the nuances of some of the struggles its users are facing.
“One thing I’ve really seen with young adults is that [AI companions] give them what they want to hear” district mental health specialist Drew Baker said. “They’re very affirming, and that is part of therapy, but the therapist’s job is also to say ‘here are these patterns that you’re doing’ and being able to identify them with continuous things. [AI chatbots] can’t identify like that.”
According to APA Services, “unlike a trained therapist, chatbots tend to repeatedly affirm the user, even if a person says things that are harmful or misguided.” This can lead to dangerous situations presented by the user which the chatbot encourages.
“I do have concerns about lonely or depressed teens [or any individual] seeking out companionship/relationship with AI as if it is a real person they are connecting to,” school counselor Brandi Cosgrove said. “I see real danger in that.”
“In two cases, parents filed lawsuits against Character.AI after their teenage children interacted with chatbots that claimed to be licensed therapists. After extensive use of the app, one boy attacked his parents and the other died by suicide.”
AI isn’t always able to recognize the difference between healthy habits and dangerous ones.
“That’s very hard. That’s affirmation, right? It’s part of it. But the other part … is saying ‘that sucks [that] you’re going through that, but here are the things you’re doing to lead to that,’ and [saying] things like that in a caring manner,” Baker said.
“I think we talk in terms of safety when we don’t have a person who is … experiencing suicidal thoughts or harming themselves, [the AI] can give resources, but there is no process to really determine if those accusations are something they follow through on. … an AI might have the 988 number [suicide hotline] but do we know if the student is going to use those? Do we know what the fallout looks like? … Versus, a person in practice is going to [say] ‘okay, let’s assess this. Do we need to call safety or is it safe to do something else instead?’”
Fortunately, it seems most teenagers are at least partially aware of the differences between a trained professional and a chatbot. Though 71% of teens have used an AI companion, according to Common Sense Media, only 12% of them use it for mental health support.
Being intentionally vulnerable isn’t the only threat. Using chatbots merely as a companion poses dangers of its own.
“Typically, when people have friends, they can check on each other if one says something out of life, is offensive, and/or dismissive,” Cosgrove said. “With AI, we do not know if the program would change to encourage someone to isolate, cut off other friendships, not take care of their mental health, etc. There is no monitoring of safety.”
“Not that all people are monitored or safe companions for everyone, but there are typically signs that something may be wrong. I am just not comfortable with it because it is not a human person.”
However, there are more dangers to AI therapy than just the uneducated advice it provides. Terms of service agreements for many of these companies include rights to collect, use, or sell personal information shared by the users with the chatbots. While these terms are provided to the users, the lack of age-verification provides children and teenagers with the opportunity to use these chatbots without understanding what personal information they are providing or if the content will be age-appropriate.
“They are taking advantage of people not paying attention to their data being used,” Cosgrove said.
For example, according to Common Sense Media’s study, Character.AI‘s terms [as of June 2025] grant the company “a nonexclusive, worldwide, royalty-free, fully paid up, transferable, sublicensable, perpetual, irrevocable license to copy, display, upload, perform, distribute, transmit, make available, store, modify, exploit, commercialize, and otherwise use the Content for any Character.AI-related purpose.”
As people use AI as a makeshift therapist, they provide very sensitive information about the struggles they may be going through or their personal thoughts. The nature of these licenses means that personal information shared by teens—including intimate thoughts, struggles, or personally identifiable information—can be retained, modified and commercialized indefinitely, even if teens later delete their accounts or change their minds about sharing.
There are a number of alternatives students have available to them that are more regulated than AI when it comes to mental health services. Depending on the needs of the student they could contact a friend, a therapist, one of the school counselors, Baker, or even reach out to a hotline. Anything with a human being on the listening end is safer than using AI for sensitive situations.
“Counselors, whether therapists or school counselors, abide by a code of ethics that AI would not be subject to.” Cosgrove said. “It is being updated this year to be ACA Code of Ethics 2025.“
“There’s no risk assessments or anything like that,” Baker said. “You may have conversation building with AI but the AI is just generating information from the environment and then spewing out a conversation piece.”
Safe ways to use AI will be different for everyone, but vigilance is important in any use case. Keeping in mind the data collection policies are important when interacting with a chatbot. While it seems quite dangerous to solely use AI for mental health support, it may be safely used to supplement other types of support.
“I’ve used it myself to kind of just remind me of some things I learned in therapy and being able to put it back into practice. I think of good coping skills and things like that. But it’s hard to replicate the knowledge that’s learned in school.” Baker said. “[The chatbots restate] exactly what they’re getting from the internet, and it’s hard to verify those sources. But, I think it’s a great tool for reminders.”
“Maybe ask AI ‘the best ways to support my mental health’, ‘healthy coping strategies’, ‘how to lift my mood through exercise’, ‘activities’, ‘ways to connect with others in person’, etc.” Cosgrove said.
Change should be made to ensure the safety of young people and their mental health. Everyone has the opportunity to make a difference in the safety of how AI is used.
“[Students should] be more aware of the pros and cons of AI and Understand how to use AI safely. What signs to look for to determine if AI is unsafe? Making sure not to use AI to hurt themselves, others, or to deceive others” Cosgrove said.
“[For parents it’s the] same as the students, but under the lens of a parent. Pay attention to what apps students are using and talk to their children about the mental health and companionship piece.” Cosgrove said. “Try to keep an open dialogue with their children so they know that they can come to them about anything and that together they can figure out a plan to support their children.”
According to Common Sense Media, parents should “start conversations about AI relationships, recognize warning signs of unhealthy AI usage, learn about the specific risks for teens, [and] explain that AI companions are designed to be engaging.”
“I believe awareness and education is essential…understanding the benefits and the dangers,” Cosgrove said. “And then creating rules, laws, standards, expectations, etc. about what would be acceptable use of AI in whatever environment—government, education, healthcare, etc.—and what would be a violation. This is an ever evolving, rapidly advancing issue that will need to be constantly reviewed and revisited to make adjustments.”
