Drawing from real conversations with young people, Harkness Fellow Caroline Figueroa shares insights about the risks perceived by young people when turning to AI chatbots for emotional support.
Imagine you’re sixteen years old. It’s past midnight, your heart is racing, and you’re spiraling into a panic about a huge test tomorrow morning. You can’t catch your breath, let alone sleep. Your family is sleeping, and you don’t want to bother your friends, who have their own stress to deal with. In that lonely moment, what do you do?
These digital tools, which are always available, are becoming the first line of support. Many young people find the quick responses helpful, but the advice can sometimes range from dismissive to dangerously inaccurate. My collaborators at Hopelab and Stanford and I spoke to 48 teens and young people aged 14-22 who named this and other risks associated with seeking emotional support from AI chatbots. One teen shared a telling example of AI missing the mark, sharing:
“And I went to AI and said: I’m really stressed about this class. What should I do? I was expecting it to say, you shouldn’t give up, and motivate me. Then it told me it seems like you’re really stressed right now. Maybe you should take some things off your plate and maybe quit this class. I wasn’t satisfied” (Asian teen girl).
This isn’t an isolated case. In fact, nearly 9 in 10 young people aged 9-17 are now interacting with AI chatbots, such as ChatGPT, Claude, or Gemini, and a quarter of them use AI daily.
In 2026, about 20% of 13- to 22-year-olds explicitly seek out AI for mental health advice. New research developed in partnership with Hopelab and the Center for Digital Thriving showed that teens turn to AI because it can feel more available and less judgmental than humans. Young people told us they are making real trade-offs because human support is not consistently available to them, or because AI sometimes gives better, more personal, or concrete advice. This report builds on the previous one by exploring young people’s perspectives on the risks associated with using AI for emotional support.
Young people are not passive users of AI. The risks they name are lived and specific. Young people are already using these tools in nuanced ways, and failing to involve them in AI governance is in itself a risk. This is why we need to move into a frame of youth agency and structural accountability. The “fix” is not just to make AI safer but also to make human support more available, responsive, and less judgmental. The next move belongs to AI companies, providers, and policymakers who are in a position to act. They need to do so with young people at the table, not just as passive users or research subjects.
Footnote: Quotes have been lightly edited for length and clarity.
When asked about the core risks of AI chatbots, the following main concerns emerged: AI dependency, harmful advice, negative impacts on relationships, sycophancy, and data privacy.
One of the most significant fears among young people is the risk of becoming overly dependent on AI to navigate daily life. There is a growing concern that without these chatbots, they might struggle to solve problems, regulate their emotions, or manage interactions with peers, parents, and teachers. As one young person put it: “I don’t want to get emotionally attached. Like if the servers shut down, I’m just out of luck, and I need that not to be on my plate” (Black teen girl).
Another young person said, “I feel like it’s making me worse because I’m becoming so reliant on it to the point where it’s kind of unhealthy. And I’m really trying to stray away from AI into my human connections and the resources I have around me instead of gravitating straight to using AI for anything” (Asian teen girl).
There is also deep concern about what this dependency means for the future, particularly for younger teens still developing social and emotional skills. The fear is that we are moving towards a world where the most important emotional bonds are with AI systems.
As one teen reflected: “I imagine these young people like me are going to a world where we’re addicted to this robot. Let’s say you’ve used AI since you’re 14 and now you’re 21, and can’t function like a normal person because you’re addicted to talking to this robot, and need your phone to talk to it. Let’s say you’re graduating and the first person you tell is AI. That’s not helpful. And I don’t think it should be normalized” (Black teen girl).
Because AI is always available and tailored to the user, there is a risk that it becomes a substitute for real people, potentially leading to isolation or a loss of social skills. Some teens even feel that AI provides a way to avoid the vulnerability required in human interaction.
As one young person put it: “I think it would be probably better if I didn’t use AI, because then I might be more comfortable with, like, speaking up about my feelings, instead of just being able to, like, hide behind a robot and talk to it” (Black young woman).
We asked young people about recommendations for policymakers and AI companies, and they offered different ideas, which we divided into three categories.
Many of the young people we spoke to wanted AI chatbots to give shorter responses that provide immediate advice. They also wanted AI chatbots to be less agreeable and more challenging of the user’s opinion. This could help young people be less dependent on AI’s advice, while being more helpful.
One young person said: “I know a lot of times, like when you want real advice, like, sometimes the truth hurts and they’re like, okay, what you said was not good at all. Like, that kind of sucked. You were being a bad person. But ChatGPT is just like, I understand why you said that in the heat of the moment. Like, don’t tell me that I want you to get it. Like, give me the facts” (Black young woman).
One young person said: “Such as when you’re doing assignments or homework, have AI reach a limit of questions so that way they [youth] can figure out how to continue to answer on their own.” (Black young woman)
At the same time, some young people also wanted the AI to sound more human-like: “I feel like maybe making the conversation more human that you would hear in a text that you’re sending to your friends would definitely be a lot more helpful, at least for me” (Asian teen girl).
Young people mentioned the need for more regulations and age-based restrictions for AI. One AYA said: “It’s all about finding that balance because it’s one thing to just completely ban it, because that just gets rid of all of the positives, right?” (Asian boy, 16 years old)
At the same time, they also felt like AI was being pushed on them, and wanted more choice about whether to use AI or not.
As one young person said. “AI has got to the point where it’s not that much autonomy. It just pops up anywhere, at any time. And it’s like, I can’t really control it because it’s like AI. But like the majority of the time during my day, I do not want to see it” (Black young woman).
Young people are the experts in how they use technology. They also have the most to gain and lose from AI developments. They want technology designed for them, and with them.
Young people pointed out that AI chatbots often lack genuine empathy. For example, one young person who used AI for emotional support said:
“It wasn’t a real person. And it would just hit me, sometimes a bit harder that this is a chatbot. This is really sad” (Hispanic non-binary teen).
The stakes can be high. We heard alarming stories about young people receiving potentially dangerous health advice, like AI suggesting severely restrictive calorie counts when asked for healthy diet tips.
Also, in day-to-day moments of stress, AI isn’t always helpful. As one teen shared:
“I was asking Snapchat AI and I was like, I’m feeling very anxious, very stressed. My heart beat is really fast. What can I do? And so it gave me a list of things that I could do, like get up and maybe try to breathe. But when I was, when I was trying to breathe and trying to calm myself down, it wasn’t really working” (non-binary Asian teen).
Many young people are also worried about the potential for AI to cause severe harm, with many noting fears about the role these chatbots might play in escalating mental health crises. One young person said: “I know that it’s bad to talk to AI and ask personal questions. I think there have been multiple cases, like the case with that boy. AI, pushed him to kill himself” (multiracial teen boy).
Many of our young interviewees were concerned that AI chatbots over-agree with users. By validating people’s perspectives and ideas regardless of whether they are right or wrong, AI chatbots may encourage ‘confirmation bias’ (a tendency to search for or interpret information that supports one’s prior beliefs or values), being overly agreeable with users, and prevent people from engaging in necessary, challenging conversations with others, or changing unhelpful behaviors.
As one young person put it: “It’s a ‘yes ma’am’ chat where it just feeds into your own delusions, I would say. And it doesn’t really give critical advice. And what I need is critical advice” (Black teen girl).
Because chatbots often side with the user, they can escalate minor misunderstandings into major conflicts. One young adult shared: “But, also, when I told it about very minor conflicts or stuff that just kinda upset me a little bit, also it would jump straight to, like, he’s mistreating you” (Hispanic young non-binary person).
Many of the young people we spoke with expressed concern about how large AI companies handle their sensitive personal data.
“I’m always scared that this information could get out and that wouldn’t be good for me at all. So I feel like that is one of the downsides, just because while you do share a lot and you do trust AI a lot, there’s always a chance that, you know, this information could get out, especially because you were like putting it in writing and you’re giving it to basically another company or another entity to use that and analyze it” (Asian teen girl).
Some also remarked that they don’t trust AI companies; they recognized that AI is not designed for user well-being, but their business model is designed to collect as much data as possible, and that AI companies may collaborate with bad actors.
One young person said: “It’s good to be cautious because, especially with everything that’s going on nowadays, you never know if, for example, these AIs or companies would cooperate with people who don’t have the same values” (Hispanic young man).
As one young person in our interviews said: “I think they [AI Companies] need to create separate AI chatbots for young people. I don’t think that they should be the same as just a normal chatbot. And I also think that they need to talk to young people and understand their psychological minds to just see how AI could not be dangerous and how AI could help them” (Black teen girl).
