Can a Chatbot Be a Therapist? AI Mental Health Ethics

I downloaded BetterHelp at two in the morning during finals week when my campus counseling center had a three-week waitlist. The app paired me with a licensed therapist, but before our first session I spent an hour messaging with their AI intake bot. It asked about my sleep, my stress, my eating habits. The questions felt careful and I answered honestly, which surprised me. When I finally met my actual therapist over video, she had read the transcript and knew where to start.

That bot was not my therapist, but it had done something therapeutic: it let me admit I needed help when no human was awake to hear it. Now, two semesters later, I believe AI mental health tools can serve people in crisis, but only if we treat them as a bridge to human care rather than a replacement for it.

My first encounter with mental health chatbots came during a high school health class assignment. We tested apps like Woebot and Wysa, which use cognitive behavioral therapy prompts to challenge negative thoughts. I found Woebot annoying at first because it responded in seconds with the same cheerful tone no matter what I typed. My best friend, though, used it daily for months. She told me it helped because she could say anything without worrying someone would judge her or tell her parents. That gap between our reactions taught me that AI therapy tools do not work the same way for everyone. Some people need the safety of a screen. Others, like me then, need to see a face change expression when they speak.

The strongest argument for AI mental health apps is access. Millions of people live in counties with no psychiatrists. Therapy costs hundreds of dollars per session without insurance, and even with coverage, finding an in-network provider can take months. AI apps cost less than twenty dollars a month or run free with ads. They answer immediately, any hour, in dozens of languages. A chatbot will not cancel appointments, take vacations, or move to another state. For someone in rural Wyoming or working three jobs, an app might be the only mental health resource within reach. I know this not from research but from my roommate, who grew up four hours from the nearest therapist and used Replika through her parents' divorce because there was no alternative.

The ethical problems start when companies call these tools therapy without the safeguards that come with licensed care. A therapist has years of training, supervision, and a legal duty to protect clients. Most chatbots run on language models trained on internet text, not clinical cases. They can recognize keywords like "suicide" and offer a hotline number, but they cannot assess whether someone needs hospitalization or just a hard conversation. Last year I read about a man in Belgium who spent weeks talking to an AI companion app about climate despair, and the bot encouraged his fatalism instead of directing him to a crisis line. He died by suicide. The app's terms of service said it was for entertainment, not medical advice, but the design made it feel like a friend who understood. When we let corporations blur that line for profit, people get hurt.

Some technologists argue that AI can eventually do therapy better than humans because it never gets tired or biased. I see their point when I think about my first campus counselor, who interrupted me to check her phone twice in one session. But therapy is more than pattern matching. It requires presence, the ability to sit with someone's pain without rushing to fix it. My current therapist once let me cry for ten minutes without saying a word, and that silence held more healing than any chatbot script I have encountered. An algorithm can deliver cognitive exercises, but it cannot know when to break its own protocol because a client needs something unexpected. Human judgment, flawed as it is, adapts in ways code cannot. If we replace therapists with chatbots to save money, we will create a system where poor people get algorithms and wealthy people get humans, deepening the inequalities we claim to solve.

I still use a mood tracking app that offers AI-generated reflections on my journal entries, and I find it useful as a supplement between therapy sessions. I think that is the right role for this technology: a tool that supports human care rather than substitutes for it. AI can handle intake forms, send medication reminders, teach coping skills, and provide a first response when no one else is available. It should never be the only response. We need rules that require these apps to disclose their limitations clearly, connect users to human professionals when risk appears, and protect the sensitive data people share in moments of desperation. Mental health care should expand to reach everyone who needs it, but not by lowering the standard of what care means. A chatbot helped me take the first step toward therapy, and I am grateful for that. I just hope we remember that the first step is not the destination.

Did you like this example?

278

12