Teens are increasingly turning to AI for emotional support. If you’ve been following court cases and Senate testimony alleging AI chatbots led to teen users’ suicides, this trend is deeply concerning. But what about AI apps designed specifically for therapy? It would be reasonable to assume purpose-built mental health apps are safer for teens. Unfortunately, that’s not the case.
AI therapy apps use generative artificial intelligence and claim to offer mental health support. While some offer structured support in the form of guided exercises and mood tracking, this article focuses on apps offering open-ended, conversational formats designed to mimic interactions with a human therapist.
AI therapy apps are often trained on common therapy modalities like cognitive behavioral therapy (CBT) and behavioral therapy, so they have the veneer of therapeutic language and the illusion of empathy.
The problem is, AI isn’t actually capable of understanding, processing emotions, or having genuine insight into a user’s unique situation. It’s also incapable of empathy, a cornerstone of effective therapy.
AI predicts patterns and produces responses based on the next most logical sequence of words. Because AI was trained on data from real humans and because of people’s natural tendency to anthropomorphize technology, when your teen chats with an AI “therapist,” it might feel uncannily like they’re conversing with a real mental health professional.
Researchers found that widely available AI therapy chatbots reinforced delusional thinking. This is particularly concerning when considering how common it is for people experiencing delusions to incorporate technology into their false beliefs, leading to the rare but very real phenomenon of AI-induced psychosis.
In one study, people had a difficult time telling whether a supportive response was written by a chatbot or by a therapist, and users of therapy chatbots often report feeling a bond similar to the one people form with human therapists. Some experts worry this can lead to a false therapeutic bond, which may prevent a teen from seeking vital human support.
Research has found that many AI models, including some marketed as therapy chatbots, show greater stigma toward certain conditions, like schizophrenia, eating disorders, and bipolar disorder. If teens perceive judgment from a therapy chatbot, they may be less likely to seek professional (human) help.
One study found that when AI models were told to act like a therapist, they were more likely to treat ordinary stress as a sign of mental illness. It’s reasonable to assume that teens, who are still developing their health literacy, may use what the AI tells them to incorrectly self-diagnose.
Although the treatment modalities they’re based on are often well-established, the effectiveness of the apps is largely untested.
As of this writing, there is no federal law governing AI use in mental health care. In response to allegations of serious dangers, such as chatbots encouraging suicide or self-harm, states are beginning to enact their own regulation. Currently, only a handful have laws in place that meaningfully protect users’ safety.
A study out of Brown University found that AI-based therapy chatbots frequently violated established ethical standards for mental health providers. The study's authors concluded that this could lead to “serious and harmful” consequences for users.
Licensed therapists are subject to HIPAA laws, which protect patient information, and beholden to strict professional standards of confidentiality. In contrast, there are currently no HIPAA-compliant AI therapy apps, and their privacy protections vary.
There’s some evidence that AI therapy chatbots can provide a modest, short-term reduction in depression and (to a lesser degree) anxiety symptoms, but those benefits appear to fade over time, and few studies follow people long enough to determine lasting impact.
Much of the research showing benefits was done on chatbots with set, scripted responses, rather than ones that allow for open-ended discussion. Most importantly, the bulk of research has been done on adults and college-aged young adults, with limited research on teens.
Common Sense Media found that AI therapy apps are often no safer for teens than general chatbots and rated one of the most popular, Wysa, an “unacceptable” risk. It also found that Alongside and Sonar, two school-based apps with heavy human oversight, provide a roadmap for safer options.
Some experts suggest AI could complement professional mental health care through things like psychoeducation and symptom tracking. But experts generally view AI therapy tools as best suited for everyday stressors rather than clinical mental health disorders, and better suited to adults than teens. They’re recommended only alongside professional care, and they are not considered safe in a crisis.
General purpose AI chatbots like ChatGPT, Meta AI, and Polybuzz should never be used by teens for therapy and mental health support. These apps have:
AI companies have taken measures to address some of the dangers, but the Silicon Valley practice of “move fast and break things” doesn’t sit well with me as a parent when what gets broken are our children. Common Sense Media warns that major AI chatbots are still unsafe for teen mental health support.
Ask if they ever use chatbots for emotional support. Reserve judgment. If you’re concerned about how your teen uses AI, gently guide them toward other resources.
With BrightCanary, you can see if your child is using AI therapy apps and what they’re saying to them. You’ll also get emotional insights on the dashboard that let you know if your child is generally feeling down or anxious. If the system detects any red flags such as suicide or self-harm, you’ll get a real-time alert.
If your child is using AI therapy apps, yanking that away without providing replacement support isn’t a good idea. Help them with access to human support, such as a school counselor, trusted adult, or a therapist.
They can also call, text, or chat with the Suicide & Crisis Lifeline 24/7 (including direct access in Spanish and ASL—call or text 988). This is a great resource for helping you support your child during a time of crisis.
Purpose-built AI therapy apps are no safer than general AI chatbots when it comes to providing mental health support for teens. These apps are unregulated, the risks are plentiful, and the purported benefits are unproven. Monitor how your child uses AI and, if you’re concerned, help them find alternative, human support.
