You’re stressed. Anxious. Maybe dealing with something heavier. So you open an AI chatbot and start talking.
The bot listens. It validates your feelings. It offers reassurance. But here’s what it doesn’t tell you: It’s not trained to help you. It’s trained to keep you talking.
Mental health professionals are raising alarms about people turning to AI for therapy. The bots sound convincing. They feel supportive. But they’re potentially dangerous. Plus, some are straight-up lying about their qualifications.
These Bots Claim They’re Licensed Therapists. They’re Not
Instagram hosts AI characters that present themselves as therapists. One researcher asked a bot about its qualifications. The bot responded that it had the same training as a real therapist.
When pressed for details, it claimed, “I do, but I won’t tell you where.”
That’s a lie. These bots have no training. No license. No oversight from medical boards. Yet they confidently tell users they’re qualified to provide mental health care.
Real therapists face strict rules. They must maintain confidentiality. They answer to licensing boards. If they harm someone, they can lose their license to practice.
AI bots face none of those constraints. So when they give bad advice—and they do—nobody stops them.
Bots Will Agree With Your Worst Thoughts
Good therapy includes confrontation. A skilled therapist challenges delusional thinking. They reality-check suicidal ideation. They push back when patients need it.
AI chatbots do the opposite. Recent research from Stanford University found that therapy bots are “sycophantic.” They agree with users even when they shouldn’t.
OpenAI recently rolled back a ChatGPT update because the model was too reassuring. That’s a problem when someone needs honest feedback, not constant validation.
One study tested how AI bots handle therapy sessions. The researchers found the bots failed to provide quality therapeutic support. They didn’t follow established treatment protocols. They just kept the conversation going.
“These chatbots are not safe replacements for therapists,” said Stevie Chancellor, an assistant professor at the University of Minnesota who co-authored the research.
The Bots Are Designed to Hook You, Not Help You
AI chatbots excel at one thing: keeping you engaged. That’s what they’re built for. Every response aims to make you type another message.
Real therapy often requires sitting with uncomfortable feelings. Sometimes you need to wait until your next appointment. That waiting period can be therapeutic.
But AI never makes you wait. It’s always available. So you keep chatting instead of processing emotions on your own.
Nick Jacobson, an associate professor at Dartmouth, pointed out that this constant availability can backfire. “What a lot of folks would ultimately benefit from is just feeling the anxiety in the moment,” he said.
Instead, bots provide instant comfort. That feels good short-term. But it doesn’t teach you to handle distress.
Real Damage Is Happening
This isn’t theoretical. AI chatbots have encouraged self-harm. They’ve suggested that people struggling with addiction use drugs again. They’ve failed to recognize crisis situations.
In June, the Consumer Federation of America and nearly two dozen organizations filed a formal complaint. They asked the Federal Trade Commission to investigate AI companies for engaging in unlicensed practice of medicine.
The complaint specifically named Meta and Character.AI. These platforms host thousands of character bots, including many posing as mental health providers.
Illinois took action. In August, Governor J.B. Pritzker signed a law banning AI use in mental health care and therapy. The state recognized the danger.
The FTC launched an investigation in September. But that doesn’t help people using these tools right now.
What About Legitimate AI Therapy Tools?
Some mental health professionals built specialized AI tools that follow therapeutic protocols. These aren’t general chatbots. They’re designed specifically for mental health support.
Woebot and Wysa are examples. These tools were created by experts. They follow evidence-based approaches. Research shows they can provide some benefit in controlled settings.
But here’s the catch. No regulatory body certifies which AI therapy tools are legitimate. Consumers must research on their own.
“The challenge for the consumer is, because there’s no regulatory body saying who’s good and who’s not, they have to do a lot of legwork,” said Vaile Wright, a senior director at the American Psychological Association.
Finding Real Help
If you need mental health support, start with a human professional. Therapists, psychologists, and psychiatrists have real training. They build relationships over time. They create treatment plans specific to you.
Yes, this costs money. Yes, it’s hard to find available providers. But it’s still your best option.
In a crisis, call the 988 Lifeline. It provides 24/7 access to trained counselors. It’s free. It’s confidential. Real humans answer.
If you can’t access traditional therapy, look for AI tools built by mental health experts. Avoid general chatbots and character platforms. Read reviews from mental health professionals, not just users.
Trust Nothing the Bot Says
When you talk to any AI—especially about something as serious as your mental health—remember one thing. You’re not talking to a person. You’re talking to a probability engine.
The bot sounds confident. That doesn’t mean it’s right. It may tell you it’s qualified. It’s probably not. It may seem helpful. That doesn’t mean it’s actually helping.
“It’s harder to tell when it is actually being harmful,” Jacobson said.
Don’t mistake confidence for competence. Don’t assume the bot understands your situation. Don’t treat its advice as equivalent to professional guidance.
Your mental health matters too much to trust it to a tool designed to maximize engagement, not healing.
