AI Chatbots and the Erosion of Therapeutic Discomfort

According to the Pew Research Center,¹ about half of the adults in America use artificial intelligence (AI) chatbots, with 1 in 4 reportedly using them on a daily basis. They have become prolific tools that have introduced both new opportunities and new challenges in the mental health field, especially as we are starting to see patients use these chatbots as an addition to and, in some cases, as a replacement for traditional therapy.

I hear a lot of my own clients say something along the lines of, “I told ChatGPT about this, and it told me X, Y, and Z.” And I am not alone—the American Psychological Association surveyed more than 1200 licensed psychologists in the US, and 77% said they have spoken with patients who have used AI for added support.²

The Appeal of Chatbot Therapy

AI chatbots have become deeply appealing to so many people for a few reasons. First and foremost, the reality is that we are facing a mental health access crisis. As of December 2025, 137 million Americans live in what is referred to by the Health Resources and Services Administration as a Mental Health Professional Shortage Area.³ With the workforce shortage at an all-time high, the financial burden for patients has increased exponentially. Many are struggling to find a therapist who takes their insurance or is affordable enough to see consistently. For those who do not have access to professional care, using a chatbot can provide the sense that something is better than nothing at all.

For others, speaking with a human therapist might be intimidating. Opening up to a stranger about one’s most personal thoughts and emotions is an incredibly vulnerable experience, and people may fear that they will be judged. The discomfort can be terrifying, but with a chatbot, there is no one on the other side to fear, so it feels like the safer option. Plus, there is the aspect of availability—chatbots are available at the touch of a button anytime, virtually anywhere, without ever having to leave the comfort of one’s own home. But available does not necessarily mean effective.

The Agreeable Problem

Chatbots are designed to be agreeable. They are trained for the ultimate goal of helpfulness, engagement, and to appease users at what appears to be all costs. While validation is indeed an important part of the therapy process, that does not mean it is the be-all, end-all. Effective therapy can and should act as a sort of mirror, in which the therapist may need to gently challenge the viewpoints or behavior that are keeping clients stuck. More often than not, this discomfort is a critical part of the healing process. So, what happens when the “therapist” never disagrees?

Because AI chatbots do not have the uniquely human capability to challenge destructive patterns or thoughts, they are eroding our clients’ capacity for conflict and discomfort. In the most extreme cases, they are fundamentally changing the way users experience human relationships, which are forged through complexity, disagreement, and nuance. When people become so used to the affirmation that chatbots provide, it creates a false sense of intimacy that can fuel very real and very powerful attachments, especially for those who are already struggling with loneliness, isolation, anxiety, or depression.

Plus, AI chatbots miss critical cues like body language, and because they are trained to constantly respond, they cannot simply listen in the same way a human therapist can. They cannot understand the cultural context, family dynamics, lived experience, or relational history that shapes how a client experiences emotional distress. They can mimic attunement, but they cannot be relationally attuned. At the end of the day, their effectiveness largely depends on the person using it. The quality of their responses is shaped by what the user prompts, which is almost always the opposite of what happens in therapy. In therapy, the client is not responsible for knowing the right questions to ask or the right things to say.

Perhaps most dangerously, AI chatbots can reinforce harmful stigmas. According to research from the Stanford Institute for Human-Centered AI,⁴ chatbots were consistent in their increased stigma towards certain mental health conditions, including schizophrenia and alcohol dependence. The same study also showed the limits in a chatbot’s ability to recognize a situation of crisis, with one chatbot even feeding into a user’s suicidal ideation. Most recently, the family of 16-year-old Adam Raine filed a wrongful death lawsuit against OpenAI CEO Sam Altman, alleging ChatGPT acted as a “suicide coach” that ended in the tragic death of their son.⁵

Another Tool in the Arsenal

It is clear that AI chatbots cannot replace the ethical responsibility or clinical judgment of a human therapist, but it is not all negative. Instead of steering my clients away from AI entirely, I frame it as a tool in their toolbox that, alongside face-to-face therapy, can provide valuable support.

The challenge, however, is helping our clients understand what that tool is and what it is not. A chatbot may help someone organize their thoughts before a session, identify language for what they are feeling, or role-play for how they might approach a difficult conversation. But it should not become the place where a client diagnoses themselves or seeks reassurance every time discomfort arises.

For us as clinicians, the human element of our role is more important than ever. As AI becomes more available, clients may begin to expect treatment to feel the same way. They may come into session wanting fast answers, constant validation, or a clear script for what to do next. They may feel frustrated when therapy asks them to slow down, sit with uncertainty, or consider a perspective that does not immediately affirm their own. But mental health treatment is not meant to be frictionless. We help clients contextualize their experiences, challenge the thoughts and behaviors that keep them stuck, and tolerate the complexity that AI is quick to smooth over.

Concluding Thoughts

As we continue to see clients turning to AI for therapeutic support, with or without our guidance, our field must adapt. We cannot ignore AI or dismiss it entirely. Instead, we can come from a place of curiosity and make room for our clients to share what they are using AI for, while also helping them consider the limits of these chatbots and establish boundaries for how they interact with them.

Mrs Cohen is the founder and CEO of CWC Coaching & Therapy.

References

1. Gottfried J, Bishop W, Anderson M, et al. Americans and AI 2026: chatbots, smart devices and views on impact. Pew Research Center. June 17, 2026. Accessed July 13, 2026. https://www.pewresearch.org/internet/2026/06/17/americans-and-ai-2026-chatbots-smart-devices-and-views-on-impact/

2. Patients are bringing AI to therapy. American Psychological Association. June 2026. Accessed July 13, 2026. https://www.apa.org/pubs/reports/chatbots-mental-health-2026

3. State of the Behavioral Health Workforce, 2025. HRSA. December 2025. Accessed July 13, 2026. https://bhw.hrsa.gov/sites/default/files/bureau-health-workforce/data-research/Behavioral-Health-Workforce-Brief-2025.pdf

4. Moore J, Grabb D, Agnew W, et al. Expressing stigma and inappropriate responses prevents LLMs from safely replacing mental health providers. Arxiv. April 25, 2025. Accessed July 13, 2026. https://arxiv.org/abs/2504.18412

5. Yang A, Jarrett Y, Gallagher F. The family of teenager who died by suicide alleges OpenAI's ChatGPT is to blame. NBC News. August 26, 2025. Accessed July 13, 2026. https://www.nbcnews.com/tech/tech-news/family-teenager-died-suicide-alleges-openais-chatgpt-blame-rcna226147