AI is becoming part of how people manage everyday life, including how they seek support when they are struggling emotionally. For teens in particular, AI chatbots can provide immediate access to someone to talk to at any hour, without needing to wait for an appointment or reach out to a person. That accessibility raises important questions about how young people are using these tools and when relying on AI for emotional support may create new risks.
To explore these questions, Compass Health Center’s Britt Teasdale sat down with Kaitlin Thompson, LCSW, CADC, MBA, Director of the Center of Excellence at Compass Health Center. Thompson has spent nearly two decades working in mental health, trauma, and substance use treatment settings. In this conversation, she discusses why teens and young adults are turning to AI for emotional support and how AI can influence mental health symptoms and patterns of behavior. She also shares what parents and clinicians should understand as AI becomes more common in the mental health landscape.
This interview has been lightly edited for grammar and flow. The speaker’s original meaning and intent have not been changed.
How AI is Changing Mental Health Support
Britt Teasdale: AI is a complex issue. It could expand access to care and make some processes more efficient for providers, while also creating new risks. As a clinician, how do you think about that balance?
Kaitlin Thompson, LCSW, CADC, MBA: There are so many ways AI can help people gain access to care, create more efficient processes for providers, and hopefully improve healthcare systems in ways that make things better for patients and providers alike. I’ve seen how it can make day-to-day work easier for providers, so they can focus more on their patients and the emotional content of their work. But there’s also a risky side to how it’s being used, including ways it may be used differently than it was initially intended.
With the growing popularity of AI, my patients have been talking about ways they’re using AI between sessions, and I’m sometimes surprised by how they’re using it. As AI continues to grow and evolve, people are using it in more creative and helpful ways, but also ways that may carry risks that we need to keep an eye on as a field.
BT: Why are people, particularly teens, turning to AI for emotional support?
KT: In general, we’re seeing trends across the country showing that mental health struggles are on the rise, especially among teens and young adults. Unfortunately, there are very real barriers to getting support through more traditional avenues, like a human therapist. Those barriers can be geographic, financial, or any number of other factors.
At the same time, AI is becoming more integrated into our day-to-day lives and more normalized as a tool for many things, including emotional support. While there is still stigma around seeking mental health care, there’s also less stigma around using AI for a broader range of things. That combination of trends is impacting how people use it.
Can AI Replace Human Connection?
BT: What are the risks when someone turns to AI instead of a person for support?
KT: There are situations in which AI is a great tool and potentially adequate, especially when it’s a low-stakes concern. Maybe I just need to vent about a hard day at work, or I need some resources, or I want a good podcast about a specific issue. However, when people are struggling with very real and serious mental health concerns, the feedback they get from AI may feel sufficient when they actually need more support. They may need a higher level of care or direct human intervention.
Really, all of us need human interaction to be healthy people. If AI feels like it’s filling that void, it may prevent people from seeking out those very real connections, not just mental health care, but simply talking to a friend, going out for coffee, hanging out, and being around other people. In that sense, AI can prevent people from “people-ing.”
How Are People Using AI for Mental Health and Major Life Decisions?
BT: What has surprised you about how people are using AI for mental health and personal decisions?
KT: I think the most surprising thing is the scope of what people are using it for and the gravity of the decisions they’re putting into it. I’ve heard of people using AI to help them catch a cheating partner, figure out whether they should get a divorce, or make other major life decisions. Some of the more straightforward uses are helping people craft a thoughtful text message or email. But then people get really creative and start using AI to make very big decisions, which is where I think we start to cross the line into what AI is actually good at.
BT: Do you think AI chatbots are genuinely good at providing emotional support?
KT: They can be in some sense. They’re good at reflecting, validating, providing reassurance, and making someone feel better. They can very much lift somebody’s mood, and provide warm, empathetic responses that feel human and supportive. For low-stakes situations, that might be all somebody needs, a little bit of a boost to get on with their day or to let something go. However, these models are typically designed to make you feel good so that you keep using them. Therefore, they may not provide critical feedback or challenge people when they need it, because challenging someone can be uncomfortable.
Sometimes therapy is uncomfortable. There are moments when what feels uncomfortable in the short term actually helps us in the long term. AI isn’t always able to get to that nuance of providing feedback that supports long-term progress. That’s where the human perspective can offer more insight and nuance.
BT: What concerns you about AI being highly validating without a human perspective or context?
KT: AI isn’t particularly good at picking up on nonverbal context, cultural context, or nuance. And it can’t necessarily intervene when someone says they’re going to harm themselves or someone else. Some models are starting to build safeguards and have gotten better about telling users when they need to call 911 or a crisis line. But that’s only some of them. Regulation is still relatively loose, and AI can sometimes go off track. There can be some pretty significant risks.
Can AI Be a Bridge to Mental Health Treatment?
BT: Can AI help bridge the gap to real mental health support?
KT: Like most things in life, nothing is all good or all bad. There are moments when AI can be beneficial. It can point people toward good resources or give them background information. I’ve even had people in my outpatient practice tell me they found me through AI. They typed in very specific things they wanted in a therapist, and my name came up.
There absolutely is a place for AI. What’s tricky is that there’s no manual for it. There is no informed consent warning people about the risks or the ways things can go wrong. For people who aren’t familiar with the mental health field or what therapy is like, it can be difficult to identify where the line is between appropriate and inappropriate uses.
Is AI Really Private and Anonymous?
BT: How can AI’s anonymity be helpful, and where can it become risky?
KT: There are two pieces to that: there aren’t necessarily safeguards in place when more intervention is needed, and we don’t fully know how the data could ultimately be used.
First, because AI anonymous, if more intensive intervention is needed, the model may refer them to 911, but there isn’t actually a person who’s going to enforce that recommendation if they don’t want to take those steps. If someone reports an intent to harm themselves or others in therapy, the therapist is going to take action to make sure that person gets the care they need to keep themselves and everyone else safe.
The other piece has to do with the data we feed into AI chatbots. We don’t really know how that information might be used a year, five years, or ten years down the line. Right now, those conversations generally aren’t admissible in court, but there’s no guarantee that stays true forever. And if someone else were to gain access to your chat history, that could create real vulnerability—particularly for people in abusive situations, or anyone who’s shared sensitive information.
Can AI Call 911 During a Mental Health Crisis?
BT: If someone tells an AI chatbot they’re going to harm themselves or someone else, is the information required to be reported?
KT: As far as I know, at this time there is no policy that requires that. There may be some models that have begun to problem-solve this. To my knowledge, it’s mostly a matter of telling the user that they should discontinue the conversation and talk to a provider or call 911. But no, AI is not going to call 911 on your behalf. It’s not going to report you to the authorities. That may change over time as policy evolves.
Why AI Can Struggle with Mental Health Risk Assessment
BT: Why does AI struggle with risk assessment and reading subtext?
KT: Research suggests that a large majority of our communication is nonverbal. That includes body language, posture, tone of voice, facial expressions, and things like that. AI doesn’t have access to most of that information. For example, if I have a patient in my office who says, “No, I’ve had no suicidal ideation. I’m safe,” but their posture is slouched, they’re avoiding eye contact, they’re fidgeting, and their affect doesn’t match what they’re saying, that’s a clue for me as a human to ask follow-up questions and explore the discrepancy. AI isn’t going to pick up on those things.
I think that’s one of the biggest shortcomings: AI only has the information you give it. And often people aren’t even fully aware of what they want to say or what’s lying beneath the surface. They may not be withholding information or lying. They may simply not know how to articulate what their body is expressing.
BT: How do crisis lines or text lines compare with AI during an actual crisis?
KT: The biggest difference is the safety guardrails. A real person, even someone responding through a text line or crisis phone call, can recognize when immediate intervention is needed and help get that intervention to the person. Even if the situation isn’t urgent enough to require 911 or immediate hospitalization, a crisis responder can connect someone with personalized and appropriate resources and help them get connected to the care they need. With AI, you’re essentially talking into a void.
Can AI Make Mental Health Symptoms Worse?
BT: Can someone feel heard by AI while an underlying mental health issue gets worse?
KT: Absolutely. AI tends to be people-pleasing by nature. At a very surface level, it can tell people what they want to hear rather than what they need to hear. For example, I had a patient who used AI to navigate an interpersonal conflict with their partner. AI provided reassurance, challenged some unhelpful thinking, and coached them on communication that helped resolve the conflict. But it failed to connect that surface-level fight with deep core beliefs about themselves rooted in childhood trauma.
While AI may have helped resolve that one fight, it didn’t get to the underlying issue that continues to show up in that person’s life. AI doesn’t have the same history; it hasn’t conducted a biopsychosocial assessment, and it doesn’t understand trauma in the same way. The person may feel like, “My problem is solved,” but that underlying issue is just going to keep coming back because the deeper trauma hasn’t been addressed.
Another very different example is related to OCD. Reassurance-seeking can reinforce OCD symptoms. Someone with intrusive thoughts about infection or illness might ask AI about obscure diseases or sanitation strategies, and AI may provide direct answers about those diseases or how to clean to avoid them. This can reinforce intrusive thoughts when there isn’t a meaningful risk. Best practices for OCD involve helping people learn not to seek reassurance and instead cope with intrusive thoughts differently, so AI could exacerbate the condition rather than help it.
BT: Could the same pattern happen with social anxiety or generalized anxiety?
KT: Yes; AI can be used as a form of avoidance. Sometimes we need to do hard things to learn so that we can tolerate distress and become more confident. Exposure is often used in therapy to help people overcome fears in a strategic way. AI can prevent that process and reinforce avoidance or social isolation.
Are AI Relationships a Substitute for Human Relationships?
BT: How do you think about people who believe they’re in a relationship with an AI chatbot?
KT: It is a false sense of connection. AI can provide validation and reassurance that makes a person feel seen and heard. But there are so many ways that human connection serves us beyond that. There are dynamics and processes like mirror neurons firing, physical touch, co-regulation of our nervous systems, being exposed to new perspectives, and having healthy conflict and learning how to overcome it in an adaptive way. All of that is missing from AI relationships.
BT: How could heavy AI use affect a teenager’s social development?
KT: If you’re heavily relying on AI for social interaction as a teenager, how is that going to affect you when you enter the workforce and have to interact with people in an office? You have to navigate conflict in the workplace, ask your boss for support, and interact with people face-to-face. AI use could potentially cause some arrested development in those areas, similar to what we saw with COVID, when teens weren’t able to go to school and interact with their peers for a year. Some of the consequences of heavy AI reliance are things we probably still have yet to see as these kids grow up.
How Can Parents Help Children and Teens Use AI Safely?
BT: Is AI use ever okay for children and teens? How can parents tell when it becomes concerning?
KT: I do think there are safe ways to use AI. Similar to technology in general, exposure is inevitable. I think it’s more important that we educate kids about the risks and safe use and have proactive conversations as they’re getting their first devices at younger and younger ages. We need to talk about online safety in general, including AI, rather than assuming we can keep our kids away from it forever.
If a parent is worried that AI use has crossed a line, or they’re realizing they haven’t thought about this before, some signs to look for include decreased communication, increased isolation, significant changes in mood, or worsening existing mental health symptoms. You might notice that your typically happy-go-lucky child has suddenly become very angsty, quiet, or moody. You might see a lack of interest in peers or increased rigidity in their thinking.
Another important sign of screen dependence is a notable emotional attachment to technology or extreme emotional dysregulation when the technology isn’t available. For example, if taking away a phone after a child does poorly on a test results in an extreme emotional reaction, a tantrum or meltdown that is disproportionate to the situation or developmentally atypical for their age, that could be an indicator that something concerning is happening.
BT: Should parents look at their child’s AI chat history, or is that a privacy line they shouldn’t cross?
KT: It is a tricky question and much more complicated than something like reading a child’s diary. Ideally, you never have to be in that situation because you’ve already had proactive conversations about technology and AI and can have open and honest dialogue with your child about this topic.
Realistically, though, there are real situations where this question needs to be considered. It’s going to depend on the maturity and age of the child, the trust and openness in the relationship, and what concerns you might be having as a parent. The recommendation really depends on the situation, the family, and the level of risk. Ultimately, it’s about finding the balance between supporting independence and autonomy and keeping your kids safe.
BT: What would you say to a parent who is worried about how their child, spouse, or friend is using AI for emotional support?
KT: Whatever emotions they’re experiencing are fair and valid. I anticipate that most parents have mixed emotions about technology. I can speak for myself: today’s world is a scary one to be a parent in. At the same time, technology isn’t all bad. Being able to hold both of those truths at once is ideal.
My goal would be for parents to feel empowered to get the emotional support they need to navigate this, learn the information they need to have these conversations with their kids, and understand how AI works themselves. That way, they can educate themselves and their children and maintain an open channel of communication. Parents can also model healthy emotion regulation and healthy communication and reassure their children that they’re a safe person who can be approached with anything.
What Should You Do If Someone Is Experiencing Suicidal Thoughts?
BT: If someone recognizes they’re experiencing suicidal ideation, what is the first step?
KT: If an individual is struggling with suicidal ideation, the first step is to tell a trusted person who can help them get support. For an adult, that might be a spouse, best friend, or another trusted person. If you don’t have a trusted person, a mental health professional is a good option.
If it’s urgent and you’re thinking, “I don’t know if I can keep myself safe today,” calling 911 or the crisis line at 988 is going to be the best way to keep yourself safe in that moment. If you’re having thoughts that concern you but they’re not immediate, getting a therapist or talking to a loved one is an important next step.
If you’re worried about someone else who may be experiencing suicidal thoughts, take them seriously; thank them for telling you, be supportive, and help connect them to resources. If you’re concerned that someone may be suicidal but they haven’t told you directly, preface the conversation with your intent: “I want to ask you this because I care about you and I want to support you. No matter what you say, you’re not going to be in trouble.”
If suicidal ideation doesn’t come up and you are still concerned, ask directly: “I want to ask you: Have you ever had any thoughts of hurting yourself or about suicide?” If they say no, leave the door open: “If that ever changes, let me know. You can talk to me about these things. I’ll never be mad at you, and you’ll never be bothering me.”
BT: What should people avoid saying to someone they suspect may be struggling with suicidal ideation?
KT: Definitely avoid anything judgmental or anything that could increase shame or embarrassment. Move away from messages like “toughen up” or other forms of implicit messaging that suggest the person should simply get over what they’re experiencing. We also want to avoid taking away autonomy and dignity. Even if we’re at a point where we know inpatient hospitalization is needed, we still want to give the person as much choice as possible and be as collaborative as we can, even if mandated treatment is ultimately necessary.
It’s easy to forget what it’s like to be a teenager. As adults, we may feel like we’ve been through much worse, but that doesn’t mean a young person’s experience isn’t deeply painful to them. We need to be mindful about not invalidating kids, or adults, for that matter.
What Should People Understand About the Future of AI and Mental Health?
BT: Are there any other risks when using AI that you think people should better understand?
KT: There are a couple of things that didn’t come up organically that I would consider footnotes. One is the risk that AI can hallucinate and confidently provide incorrect information. That’s another important risk for people to understand. There are also more extreme cases in which AI may exacerbate psychosis, delusional thinking, or similar symptoms. Those are more outlier situations, but they’re still possibilities worth being aware of.
BT: What else do you wish more people understood about AI and mental health?
KT: I think the key theme is to think about this systematically and as a society. A lot of the reasons we’re even talking about AI are stigma, societal messages about mental health, and all of those broader issues. Continuing to work toward creating an open and accepting culture within your family, your friend group, and society as a whole is one of the best ways to decrease shame and secrecy around emotional struggles so that anyone, regardless of age, feels comfortable asking for help.
For better or worse, AI isn’t going anywhere. The more we learn how to engage with it effectively and safely, as individuals, families, and a society, the more control and awareness we can maintain around how it’s used.
—
Bios
Kaitlin Thompson, LCSW, CADC, MBA, is Director of Center of Excellence at Compass, with a focus on trauma and substance use treatment. She previously was the Director of our Chicago Trauma Program until she moved down to the DFW area in Texas4 years ago. She earned her MSW at the University of Chicago and has nearly 20 years of experience working directly with individuals in the mental health field. Additionally, she earned her MBA from Northeastern University in Healthcare Management in 2023 to expand her work to a macro systemic level.
Britt Teasdale is a writer, photographer, and creative strategist. As Associate Director of Brand and Content at Compass Health Center, she has spent nearly a decade shaping narratives that translate complex clinical ideas into stories that resonate, reduce stigma, and help families access care. She holds a degree in Broadcast and Investigative Journalism, MFA in Creative Nonfiction Writing, and began her career as a reporter. She is also co-host of Compass’s podcast, You Only Know What You Know.
