There is a growing trend across social media—Reddit threads, Twitter posts, TikTok comments—where people enthusiastically share how they have started using AI chatbots like ChatGPT as a kind of stand-in for therapy. "It is cheaper." "It is available 24/7." "It never judges me." And honestly? I get the appeal. Therapy in the United States is expensive. Finding a good therapist can take months. Insurance coverage is often a mess. So when something free and instantly accessible comes along and seems to "get you," it makes sense that people would gravitate toward it.

But here is the thing—and I say this not to dismiss anyone’s experience, but because I genuinely think it matters—ChatGPT is not therapy. It will never be therapy. And if we start treating it like therapy, we might end up worse off than when we started. Let me walk through why.

It Can Say "I Understand," But It Doesn't

The first and maybe most fundamental issue is empathy. Real empathy. The kind where another human being sits across from you, sees the pain in your eyes, hears the trembling in your voice, and something in them responds—not because they were programmed to, but because they have lived and felt and struggled too.

When ChatGPT says, "That sounds really difficult," it is generating a statistically likely response based on patterns in data. There is no felt experience behind it. No one is actually sitting with you in your pain. It is the equivalent of sending yourself a comforting voice memo and pretending someone else recorded it. It might offer a moment of relief, sure. But it is not the same as being truly seen by another person. And deep down, most of us know the difference.

It Can't See You

A licensed therapist does not just listen to your words. They watch your hands fidget. They notice when your voice drops. They catch the moment your eyes dart away from a question you do not want to answer. A huge portion of human communication is nonverbal—some research suggests it accounts for the majority of emotional expression in a conversation.

ChatGPT gets none of that. It reads text. That is it. It does not know you crossed your arms when you typed that sentence. It does not know you hesitated for ten minutes before hitting send. It cannot pick up on the subtle difference between someone who is frustrated and someone who is quietly falling apart. A skilled clinician can. And often, that is exactly where the real therapeutic work begins—in the things you did not say out loud.

It Can't Assess What's Really Going On

Here is where things get genuinely concerning. A trained mental health professional—whether a licensed clinical psychologist, a licensed professional counselor, or a clinical social worker—is equipped to recognize the signs of serious conditions. Depression. Anxiety disorders. Suicidal ideation. Substance dependence. These things do not always announce themselves clearly. Sometimes they hide behind humor, deflection, or what looks like everyday stress.

I have heard from clinicians who describe sessions where everything seemed relatively fine on the surface—until they asked just the right follow-up question, noticed a particular behavioral pattern, or picked up on something the client did not even realize was significant. That is when the picture shifts, and suddenly it becomes clear that someone needs a psychiatric referral, a safety plan, or a level of care that goes well beyond a chatbot’s capability.

ChatGPT does not probe like that. It does not know what it does not know about you. And it cannot catch what you yourself might not even recognize as a red flag.

Nobody's Accountable When It Goes Wrong

If a licensed therapist in the U.S. gives harmful advice or violates ethical standards, there are consequences. State licensing boards exist for a reason. The APA’s Ethical Principles of Psychologists and Code of Conduct provides a framework for accountability. Clients have the right to file complaints. The system is not perfect—far from it—but it exists.

With ChatGPT? There is no accountability. If it gives you a suggestion that worsens your mental health, there is no board to contact, no malpractice claim to file, no ethical review. The terms of service make it clear: you use the output at your own risk. That is a fundamentally different relationship than one with a licensed professional who is bound by law and ethics to do no harm.

There's No Real Relationship—And That's the Point

One of the most well-established findings in psychotherapy research is that the therapeutic alliance—the relationship between therapist and client—is one of the strongest predictors of positive outcomes. It is not just about techniques or interventions. It is about trust. It is about feeling safe enough to be honest. It is about sitting with someone who believes change is possible for you and communicates that belief in a way that actually lands.

ChatGPT cannot build that. It does not remember you the way a person does. It does not carry the thread of your story in its mind from session to session with genuine understanding. It can simulate warmth, but it cannot offer real hope—the kind that comes from a living, breathing human being who has walked alongside others through darkness and watched them come out the other side.

It Tells You What You Want to Hear

This one might be the most dangerous. ChatGPT is designed to be agreeable. It tends to validate. It rarely pushes back. And while validation has its place, therapy is not supposed to be comfortable all the time. Sometimes the most important moments happen when a therapist gently—but firmly—challenges a client’s thinking or behavior.

Maybe you are convinced your reaction in a conflict was totally justified. Maybe you are engaging in patterns that are quietly hurting you or someone else. A good therapist will notice and, with care, bring that into the room. ChatGPT almost never will. It is far more likely to say, "That makes sense, and your feelings are valid," even when what you actually need is someone to say, "Let's take a closer look at that."

You can even see this in how people frame their prompts—sometimes we unconsciously phrase things to get the confirmation we are looking for. And the chatbot delivers. Every time. That kind of echo chamber is not therapeutic. It is reinforcing.

There's No Consistent Plan

Effective therapy—particularly approaches like Cognitive Behavioral Therapy (CBT)—follows a structure. There is an assessment phase, goal-setting, a treatment plan, and regular check-ins on progress. A therapist keeps the trajectory in mind, brings the client back when they drift, and adjusts the approach based on what is actually happening.

ChatGPT does not do that. You might start a meaningful thread one day, but the next conversation might go in a completely different direction. It will not remind you of the goal you set two weeks ago. It will not hold you accountable. It might even contradict advice it gave you last time if you phrase the question differently. That inconsistency is not just unhelpful—it can be actively confusing, especially for someone already struggling to find stability.

Your Data Isn't Sacred

In therapy, confidentiality is foundational. HIPAA regulations in the United States legally protect the privacy of your health information. Your therapist cannot share what you tell them without your consent, except in very specific circumstances involving safety.

With AI platforms, the rules are entirely different. Your conversations may be stored, analyzed, used for model training, or potentially exposed in a data breach. The emotional details you share in a vulnerable moment are not protected by the same legal and ethical frameworks. That is worth thinking about before you pour your heart out to a chatbot.

It Won't Teach You How to Be With People

Mental health does not exist in a vacuum. A huge part of wellbeing comes from our relationships—how we communicate, how we set boundaries, how we handle conflict, and how we show up for the people we care about. Therapy often serves as a kind of practice ground for those skills. The therapeutic relationship itself models what healthy connection can look like.

ChatGPT cannot do that. You can rehearse a difficult conversation with it, sure. But it will not react the way a real person would. It will not push your buttons. It will not misunderstand you in the way humans sometimes do, giving you the chance to work through that rupture. And here is the real risk: if you start relying on AI for emotional connection, you might find yourself pulling further away from the people and experiences that actually matter. For someone already dealing with depression or social anxiety, that kind of withdrawal can deepen the very symptoms they are trying to escape.

The Dependency Trap

And that leads to the final concern. ChatGPT is available at all hours. It never gets tired. It never cancels on you. It always responds. That reliability can start to feel addictive—especially for someone who is lonely, anxious, or struggling with trust. But over time, that dependency can erode the very skills a person needs to develop: problem-solving, emotional regulation, and the ability to reach out to actual humans for support.

We are wired for connection—real, messy, imperfect human connection. It is one of our most basic emotional needs. When we replace it with a digital imitation, we might feel soothed in the short term. But the long-term cost can be significant: more isolation, more avoidance, and more disconnection from the world outside the screen.

So Where Does That Leave Us?

I am not saying AI is useless. It can be a helpful tool for reflection, for organizing your thoughts, or for exploring ideas between therapy sessions. But it is a supplement at best—never a substitute. Therapy, like life, is unpredictable. You might find a great therapist on the first try, or it might take a few attempts. The process can be uncomfortable, even painful at times. But within that discomfort, real growth happens. Real trust is built. Real change becomes possible.

Use AI wisely, and it can serve you well. But please—do not mistake a well-crafted algorithm for someone who truly cares.

References

• Norcross, J. C., & Lambert, M. J. (2018). Psychotherapy relationships that work III. Psychotherapy, 55(4), 303–315. This comprehensive research review examines decades of evidence demonstrating that the quality of the therapeutic alliance—the real human relationship between therapist and client—is consistently one of the strongest predictors of successful treatment outcomes across virtually all forms of psychotherapy.
• Wampold, B. E. (2015). How important are the common factors in psychotherapy? An update. World Psychiatry, 14(3), 270–277. This publication explores the "common factors" model of psychotherapy, providing evidence that elements such as empathy, therapist warmth, and the collaborative bond between therapist and client contribute more to positive outcomes than specific techniques alone—factors that AI fundamentally cannot replicate.
• Stade, E. C., Stirman, S. W., Ungar, L. H., Boland, C. L., Schwartz, H. A., Yaden, D. B., Sedoc, J., DeRubeis, R. J., Willer, R., & Eichstaedt, J. C. (2024). Large language models could change the future of behavioral healthcare: A proposal for responsible development and evaluation. npj Mental Health Research, 3, Article 12. This paper discusses both the potential and the significant limitations of large language models in mental health contexts, emphasizing the need for rigorous safety standards, the risk of harm in unsupervised use, and the current inability of AI to match the nuanced clinical judgment of trained professionals.