ChatGPT is Not Therapy

Alright, everyone. The gloves are coming off today. I need to get some things off of my chest about how AI is interfering with the therapeutic process, and warn you of the potential dangers around utilizing it for your mental health treatment. The primary goal with this exercise is to provide you with insight around how large language models (LLMs) such as ChatGPT, Google Gemini, Claude, etc. operate when given mental health prompts about internal struggles and relationship dilemmas. My hope is that this knowledge of AI’s capabilities and limitations will help you feel better equipped in deciding how you would like to utilize these programs for your personal and professional endeavors.

Mock Case Study

Perhaps the simplest way to illuminate the dangers of utilizing large language models for therapy is to walk through a mock case example. Let’s say that you open up the AI program and write a prompt such as…

“I am worried that I am about to lose my job. What can I do to ensure that I remain employed?”

I am willing to bet that the very first paragraph of the program’s response will be some sort of reflection and/or validation of your current dilemma. It will likely state something along the lines that your anxiety is understandable because you are in a precarious situation and so naturally you want a viable solution to feel a sense of security.

Now, the question is why does the artificial intelligence program start its response with these feeling words rather than providing the user with an objective solution? After all, the user is asking for a plan of action, not an echo of what they have already written. If you ask the LLM to clarify its intent behind utilizing the words of affirmation, it will likely respond that human psychology researchers have shown that people react positively when they feel seen and heard by others. ChatGPT, Google Gemini, Claude, etc. are not human. Yet they attempt to emulate humanistic characteristics because the metrics show that diction prompts user engagement.

At first you might think, “Okay. Fair point. LLMs only have the capacity to mimic therapeutic skills. But does that really matter if I still personally find it helpful? After all, a tool is meant to be used. Hammer, meet nail.” The problem is that genuine human interaction is absolutely essential for positive mental health treatment to occur.

Common Factors Within the Therapeutic Process

Time and time again human psychology research asserts that what moves the needle for an individual’s self-actualization process is the relationship dynamic between clients and their counselors (Wampold, 2015). Cognitive Behavioral’s downward arrow worksheets, Gestalt therapy’s empty chair technique, Psychoanalytic’s dream analysis, the four existential givens of life, attachment styles, reflections of content, feeling, and meaning, immediacy, confrontation, words of affirmation, Socratic questioning, blah, blah, blah are all in service of fostering the therapeutic relationship itself. And no matter how hard these multi-billion-dollar corporations attempt to dress up their AI programs, at the end of the day it will always remain a façade: A fancy algorithm pretending to mimic real human connection.

Ethical Dilemmas for Large Language Models

Up to this point these LLM corporations have been given very long leashes for how their programs can respond to user prompts about their mental health struggles. However, at the end of 2025 there were “at least 10 known lawsuits against OpenAI (creator of ChatGPT) and Character Technologies (creator of Character.AI) with a wide range of allegations including wrongful death, involuntary manslaughter, sexual abuse, negligence, and product liability, among others” (Frances et al., 2026, para. 4). These large language model corporations are attempting to quell legitimate concerns about their programs by providing disclaimers regarding their limitations when certain situations arise, such as when users inquire about diagnoses and/or suicidal ideation.

But that approach is insufficient for people who are struggling with their mental health wellness. In much the same way that professional therapists are bound by laws and ethical codes, AI corporations need oversight and regulation. People need to be told upfront how LLMs operate so that they do not misinterpret both their strengths and limitations. When AI programs are given user prompts that mirror therapeutic interaction, LLMs need to clearly state that they are incapable of providing therapeutic treatment in all its forms because the research shows that human interaction is necessary for positive mental health development.

Conclusions

I am airing out my frustrations around how LLMs are interfering with therapeutic processes because I think that you deserve better. You should be treated with respect and dignity because true understanding and empathy is ultimately what will catalyze your self-actualization process. I genuinely want to support you and help you navigate this difficult chapter in your life.

Click on the link below to schedule a free 15-minute consultation with me.

‍References

Frances, A., Simpson, J. R., & Pierre, J. M. (2026, April 21). The psychiatrist's preview of legal cases against big AI. Psychiatric Times. https://www.psychiatrictimes.com/view/the-psychiatrist-s-preview-of-legal-cases-against-big-ai

Wampold, B. E. (2015). How important are the common factors in psychotherapy? An update. World Psychiatry, 14(3), 270–277. https://doi.org/10.1002/wps.20238