When Kids Turn to AI for Emotional Support: Important New Canadian Research Parents Should Know About
- The White Hatter
- 3 minutes ago
- 8 min read
Much of the public conversation surrounding youth and generative artificial intelligence has focused on school. Should students be allowed to use ChatGPT? Are they using AI to cheat? Can AI help with brainstorming? What happens to critical thinking when students outsource too much of their work? These are important questions, but new Canadian research published online on August 31, 2026, in JAMA Pediatrics (1) reminds us that something else is happening alongside academic AI use. Some young people aren’t just asking AI to explain algebra, summarize a chapter, help brainstorm an essay, or explain something they don’t understand, they are talking to it about how they feel.
Researchers from the University of Ottawa, University of Toronto, and Toronto’s Hospital for Sick Children examined data from 39,761 Ontario students in Grades 4 through 12, collected between May 2025 and March 2026 across four Ontario school boards. The average age of the students was approximately 13 years old. Researchers distinguished between what they called functional AI use, such as using generative AI for school related tasks, and affective AI use, meaning turning to generative AI for emotional support or advice. What they discovered should get our attention, approximately 21%, or about one in five students, reported using generative AI for emotional support or advice.
In a hypothetical classroom of 30 students, that percentage would translate to roughly six young people who have used generative AI in some capacity for emotional support or advice. This doesn’t necessarily mean those students have replaced parents, friends, counsellors, psychologists, or other trusted adults with an AI chatbot, the study doesn’t establish that. However, it does demonstrate that talking with AI about emotional issues is no longer some hypothetical future behaviour that parents need to prepare for. This is something we have been writing about and addressing in both our parent and student AI presentations for more than two years, based largely on what we have been hearing and seeing through our direct work with youth and teens (2). Until now, much of what we were observing was anecdotal. This new research provides important evidence based support for what we have been seeing in practice and helps move the conversation from observation and experience toward a growing body of research.
The researchers examined whether using AI for emotional purposes was associated with emotional problems among the students. They assessed emotional difficulties using a validated measure called the Ontario Child Health Study Emotional Behavioural Scales and compared students who reported affective AI use with those who did not.
The difference was significant. Among students who reported using AI for emotional support or advice, 57.7% met the study’s threshold for clinically significant emotional problems, compared with 29.2% of students who did not report using AI this way. Before statistical adjustments, students using AI for emotional support were therefore almost twice as likely to meet that threshold.
However, there is an important piece of research literacy that needs to accompany those numbers. Researchers also accounted for several other factors, including demographic characteristics, loneliness, and something psychologists sometimes describe as “mattering,” essentially the feeling that you are important and valued by other people. Even after those adjustments, using generative AI for emotional support remained significantly associated with emotional difficulties.
We believe that this is perhaps the most important part of this article for parents and caregivers to understand, because we can almost guarantee that headlines and social media posts discussing this research will eventually simplify it into something resembling, “AI Chatbots Are Causing Mental Health Problems in Kids.” and such a headline would go beyond what this research demonstrates.
Imagine researchers conducted a study and discovered that people sitting in hospital waiting rooms were considerably more likely to be sick than people sitting in coffee shops. It wouldn’t be reasonable to conclude from that association that hospital waiting rooms were making people sick. Many people are in the waiting room precisely because they were already sick. Something similar needs to be considered here.
A young person experiencing anxiety, sadness, relationship difficulties, loneliness, family conflict, bullying, identity questions, academic stress, or other emotional challenges may seek somewhere private to talk about what they are experiencing. For some young people, generative AI may now be one of those places. Why:
AI is available 24/7
Ai doesn’t appear impatient.
AI doesn’t interrupt.
AI doesn’t visibly judge.
AI doesn’t gossip with classmates.
AI doesn’t require an appointment.
AI can respond almost instantly. and,
AI system can feel less intimidating than approaching another human being.
This creates a very plausible alternative interpretation of the findings, rather than AI necessarily creating the emotional difficulties identified in this study, emotional difficulties may be one reason some young people are turning to AI in the first place.
The researchers themselves describe affective AI use as a potential marker of psychological distress, which we believe is an important way for parents to think about these findings. Recognizing that this study cannot establish causation should not cause us to dismiss the potential risks either.
Generative AI can produce responses that feel remarkably human. Modern conversational AI systems can respond empathetically, remember conversational context, mirror someone’s language, provide reassurance, and maintain a conversation for as long as someone wishes to continue talking. For a young person who is upset, isolated, embarrassed, or simply doesn’t feel ready to speak with another person, that experience can feel supportive.
There can be benefits to that accessibility. A young person might use AI to put feelings into words, practise how to approach a parent, brainstorm questions to ask a counsellor, understand terminology, or organize thoughts before speaking with someone they trust. However, there is a fundamental limitation that young people need to understand, “AI can simulate empathy without experiencing empathy.”
An AI chatbot doesn’t care about a child in the way a parent, caregiver, friend, teacher, counsellor, psychologist, doctor, coach, or other trusted person can. It doesn’t see everything happening in that child’s life. It cannot reliably understand tone, body language, family circumstances, developmental history, subtle behavioural changes, or all of the contextual information that a human being may recognize.
Most importantly, generative AI can be wrong. When the topic is the capital of a country or how photosynthesis works, an inaccurate answer may be inconvenient. When the conversation involves depression, abuse, self-harm, suicide, eating disorders, relationships, sexual health, or another serious emotional issue, inaccurate or inappropriate information can carry much greater consequences.
That is why this conversation needs to be more nuanced than either “AI is dangerous, keep kids away from it” or “AI is helpful, so there is nothing to worry about.” Neither position adequately reflects what is happening.
We found that one of the most useful aspects of this study is that the researchers distinguished between functional and affective AI use. Those aren’t necessarily the same thing. A student asking an AI system, “Can you explain this math problem differently?” is having a fundamentally different interaction from a student typing, “Nobody at school likes me and I don’t know what to do.” The first interaction is primarily informational, while the second is relational and emotional.
Interestingly, the researchers found only a modest association between functional school related AI use and emotional problems compared with the stronger relationship observed when young people used AI specifically for emotional support. This distinction should become part of how parents, educators, clinicians, and policymakers talk about youth AI use. Asking simply, “Does your child use AI?” may increasingly tell us very little. The more useful question might be, “What are they using AI for?”, that changes the conversation considerably.
One response to research like this may be for parents and caregivers to immediately want to check their child’s AI conversations. We understand this need, but we would urge caution. If young people believe that every conversation they have with technology will automatically be inspected by a parent, some may simply move those conversations somewhere else or become more skilled at hiding them. More importantly, if a young person is using AI because they are struggling emotionally, discovering that behaviour should ideally become an opportunity for connection rather than confrontation.
Instead of beginning with, “Show me what you’ve been asking ChatGPT,” consider beginning with curiosity. Ask your child whether they or their friends ever talk to AI when they are stressed or upset. Ask what they think AI is good at when giving advice. Ask what they think it might get wrong. Ask whether an AI chatbot feels easier to talk to about certain things than a person. Ask who they would go to if something became serious.
Those conversations can tell us considerably more than simply looking through a chat history. The goal shouldn’t be catching a young person using AI, the goal should be making sure they know that human support remains available when they need it.
We believe this research also reinforces something that is becoming increasingly important in AI literacy education:
Teaching young people how to write prompts isn’t enough.
Teaching them how to identify AI generated images isn’t enough.
Teaching them that AI sometimes hallucinates isn’t enough.
Young people also need what might be described as relational or emotional AI literacy, an understanding of what these systems are and, equally importantly, what they are not. They need to understand why an AI chatbot can feel supportive even though there isn’t another human being behind the conversation. They should understand concepts such as anthropomorphism, sycophancy, persuasive design, simulated empathy, and algorithmic reinforcement in age appropriate ways. They should know that an AI system agreeing with them doesn’t necessarily mean they are correct, that confident advice isn’t necessarily good advice, and that feeling understood by an algorithm isn’t the same as being understood within a human relationship. Most importantly, young people need to know when a conversation should move from AI to a human being.
Parents and caregivers shouldn’t read this research and conclude that they now need to understand every large language model, algorithm, AI companion, or chatbot their child might encounter, you don’t. What matters much more is understanding the role AI is beginning to play in your child’s life.
There is a significant difference between a young person occasionally asking an AI system for advice and a young person increasingly relying upon it as their primary source of reassurance, validation, companionship, or emotional support. Parents and caregivers should pay attention to patterns rather than panic over individual interactions.
If your child tells you they sometimes talk with AI when they are upset, resist the temptation to immediately respond with, “Don’t do that.” Instead, try asking, “What do you like about talking to it?” The answer may tell you something important.
Perhaps AI feels more private, your child doesn’t want to burden someone else, they feel embarrassed or unsure how to begin a difficult conversation, they find writing easier than talking, or they are simply curious about what AI might say. Whatever the reason, understanding why your child turned to AI can reveal something important about what they may be thinking, feeling, or needing, something that simply banning the technology would never have taught you.
The concerning finding of this research isn’t simply that approximately one in five students reported using AI for emotional support, it’s that some of the young people turning toward these systems may also be the young people who are struggling, and that should change how we respond. Instead of immediately asking, “How do we stop kids from talking to AI?”, perhaps the more important question is, “Why might this young person feel more comfortable talking to an algorithm than talking to another person?”
Sometimes there may be perfectly understandable reasons. AI is convenient, private, immediate, and easy to access. However, when an algorithm begins becoming a young person’s emotional refuge, particularly when that young person is already struggling, adults should pay attention. Not with fear, not with punishment, not necessarily with surveillance, but rather with connection.
AI may become an increasingly useful tool in the lives of young people, including potentially helping them understand and articulate emotions. However, we don’t believe it should become a replacement for the human relationships that help young people develop resilience, perspective, belonging, empathy, and genuine emotional connection.
The researchers behind this study reached a similar conclusion from a clinical and digital literacy perspective, arguing that we need to distinguish between using AI as a functional tool and using an algorithmic interface as an emotional refuge.
For parents and caregivers, we think there is an even simpler way to frame it, “Don’t just ask whether your child is using AI. Ask what role AI is playing in their life.” That may turn out to be one of the more important parenting questions of the generative AI era.
Digital Food For Thought
The White Hatter
Facts Not Fear, Facts Not Emotions, Enlighten Not Frighten, Know Tech Not No Tech.
References:
