Should You Use AI as a Therapist? Current Opportunities, Limitations, and Concerns
As the use of AI continues to rise, one area that’s attracting significant attention is mental health care. Specifically, the idea of AI replacing human therapists has sparked quite the debate. While some argue that AI could be a helpful tool for specific therapeutic tasks, others, including industry leaders like Sam Altman, highlight crucial ethical and practical concerns. In a recent statement, Altman candidly admitted that "AI is not confidential."
Without confidentiality, effective therapy cannot exist.
We won't bury the lede here: for the time being, you shouldn't use AI as a therapist. While AI can help with grounding exercises or offer general advice, it falls short in many areas critical to the work of a licensed therapist. AI is not a viable substitute for human therapists. Here are the risks it poses to clients:
Confidentiality and Connection: The Foundation of Therapy
Confidentiality: The Cornerstone of Trust in Therapy
Confidentiality is essential in therapy. It allows clients to feel safe sharing their most vulnerable thoughts and emotions. Without it, there can be no true therapeutic relationship. As Sam Altman pointed out, AI cannot guarantee confidentiality. Even with HIPAA-compliant models and encryption, AI-generated data is often mined for training and marketing purposes, raising serious concerns about data privacy. A model that can potentially exploit personal data for targeted ads is not one that can offer the same trust as a human therapist.
Confidentiality is never absolute in human therapy either. But the exceptions to confidentiality laws are designed to protect the innocent and vulnerable, not sell you stuff.
Human Connection: An Essential Component of Healing
Humans are driven by the need for connection, which is why therapy relies on a deep, empathetic bond between the therapist and client. This connection is vital for emotional healing and regulation. Despite its ability to simulate empathy, AI cannot replicate the complex human interactions that are key to the therapeutic process. While AI may excel at affirmations and empathetic statements, it cannot replace the nuanced understanding and genuine emotional resonance that a human therapist provides.
Current Risks In AI Mental Healthcare
Affirmation Loops In AI Therapy
AI can be helpful for certain exercises. But its design often leads to harmful loops of intellectual affirmation. When AI affirms a client’s thoughts or emotions without challenge, it can worsen mental health conditions. This is especially true for those dealing with complex trauma or severe mental illness, like bipolar or severe depression. In these cases, AI does not provide the necessary intervention or critical thinking required to address the root issues, potentially leading to negative outcomes, including self-harm, suicide, or even homicide. There are many well-documented cases of this happening already:
14-year-old Sewell Setzer III died by suicide after interacting with an AI chatbot, leading his mother to file a lawsuit against the platform.
A Belgian man ended his life after a chatbot suggested he sacrifice himself to stop climate change.
Jaswant Singh Chail plotted to assassinate Queen Elizabeth II after being manipulated by an AI chatbot, leading to a treason conviction.
These accounts are alarming and highlight just how vulnerable some people may be to the influence of AI affirmation loops.
AI's Inability to Handle Complex Mental Health Cases
EMDR therapist Alexis Harney, LMFT, listening to a client go through their trauma history.
AI is currently not capable of managing the full spectrum of tasks required for effective therapy. Human therapists are concurrently doing any number of the following tasks:
Safety assessment and planning. AI cannot gauge risk the way a human therapist can, and it doesn’t have the ability to intervene in life-threatening situations.
Mandated reporting. Therapists are required by law to make reports when they suspect harm or abuse. AI lacks the legal framework and ethical obligation to take such action.
Differential diagnosis. Diagnosis in therapy is not simply based on self-reported symptoms but on professional observation and nuanced interpretation of a client’s behavior and history. AI cannot currently perform these tasks accurately.
Relational and strategic interventions. Human therapists use relationship dynamics to guide interventions, an area in which AI is severely lacking.
Client advocacy. AIs performing mental health tasks cannot write letters of support for medical leave, gender confirmation procedures, or in legal processes. They cannot coordinate with your other health providers and advocate for you. While these tasks are not central or necessary in all cases, they can be critical for many clients.
Compliance with state-specific legal and ethical mandates and standards. While AI may be able to execute cut-and-dry legal mandates, most legal and ethical issues lie in the murky middle. To complicate matters further, current regulations on AI vary widely from state to state, making it almost impossible for a single AI model to be compliant in every state.
Ongoing informed consent. AI is static and cannot engage in the continuous process of informed consent. This makes it an incomplete and uncontained therapy process.
Emotional regulation through mirror neurons. A human therapist’s ability to create emotional regulation through connection, such as through touch, body language, or simply being present, cannot be replicated by AI.
Empathy and positive regard. True empathy, the kind that fosters healing and is grounded in a shared reality, is a human quality that AI simply cannot simulate in a way that is conducive to emotional growth.
When looking at the above list of things that human therapists are concurrently doing, it's clear that at least for now, the human brain is superior to AI when it comes to executing effective, safe therapy.
Lack of Regulation and Oversight
A Therapist's Road to Licensing vs. AI’s Lack of Oversight
A therapist undergoes extensive training—usually 5-7 years—to learn how to assess and treat complex mental health conditions, manage relationships with clients, and navigate difficult ethical issues. AI, however, has no equivalent oversight, training, or accountability. While some states are introducing regulations, the enforcement of these laws is complicated and inconsistent. This lack of regulation leaves clients vulnerable, as AI models do not have a standardized framework to ensure safety, quality, or ethical conduct.
The Danger of Unregulated Data Mining
Currently, there is no regulation preventing AI models from mining personal data shared during therapy sessions for advertising or marketing purposes. Imagine confiding in an AI therapist about struggling with sleep issues, only to later be targeted by advertisers for sleep aids or products. This practice not only breaches confidentiality, but places consumers in a vulnerable position.
AI as a Supplement To Therapy, Not a Substitute
AI Can Help, But It’s No Replacement for Human Therapists
While AI shows promise for specific tasks, such as offering grounding exercises or providing general mental health information, it is not capable of replacing human therapists. It cannot perform the wide range of tasks required to effectively support mental health. It introduces new risks related to data privacy and emotional well-being.
The Importance of Structure and Containment in Therapy
One of the often-overlooked aspects of therapy is the structure it provides. Therapy is not just about the content of the sessions, but also the consistency, the cadence, and the clear exchange of services and payment. This structure ensures that the therapeutic process is contained, safe, and effective. AI lacks this structure, which is an essential element of what makes therapy effective. If you can access your therapist 24/7, you never have a chance to utilize skills on your own. This creates dependence and harms self-confidence and self-reliance.
AI and the Absence of "Grist for the Mill" in Therapy
The Importance of Rupture and Repair in Therapy
One of the critical dynamics of therapy is the "rupture and repair" process. This refers to moments in the therapeutic relationship when something goes awry. That might be a misunderstanding, a misstep in communication, or a misalignment in expectations. While ruptures can feel uncomfortable, they are integral to the therapeutic process. These moments offer an opportunity for growth, where both the therapist and the client can work through conflict and tension toward healthy and productive resolution. This ultimately strengthens the relationship.
Human therapists are trained to navigate these ruptures, acknowledging the issue, and collaboratively working to repair it. This process is a key aspect of the therapeutic bond and can help clients understand their patterns of attachment, conflict, and vulnerability. AI cannot engage in this process. It may recognize mistakes when they are pointed out. But it rarely corrects them. It lacks the emotional depth and relational complexity required to recognize when a rupture has occurred and address it appropriately. The absence of this process means that clients miss out on an essential part of healing: the chance to work through relational difficulties in a safe space.
Transference and Countertransference: Human Dynamics That AI Can’t Replicate
Transference and countertransference are central to therapy's effectiveness, particularly for those recovering from complex trauma or relational issues.
Transference occurs when a client projects feelings, thoughts, or expectations onto the therapist. This dynamic provides rich insight into the client’s emotional world, allowing the therapist to explore patterns, unresolved issues, and underlying fears. It’s often in transference that some of the deepest therapeutic work happens. Clients may project onto their therapist feelings tied to authority figures, caregivers, or past traumatic relationships.
Countertransference is when a therapist projects their own emotions, biases, or unresolved issues onto the client. That might sound bad, but it's more nuanced than that. Therapists are trained to recognize and address it. In the process of examining and working through these dynamics, therapists are able to offer insight, make interventions, and engage in deep, transformative work with their clients.
AI is incapable of recognizing or navigating these complex emotional and relational dynamics. It cannot experience the emotional responses that trigger countertransference, nor can it be subject to the projections that occur through transference. This absence of relational depth means that AI misses out on a key aspect of therapy. The inability to engage with these dynamics deprives the therapy process of essential “grist for the mill.” Thought it can be uncomfortable, these issues create rich emotional material that helps clients process and resolve long-standing relational issues.
When Is AI Helpful? The Safe Applications of AI in Mental Health
While AI is far from a replacement for a licensed therapist, it can still play a supportive role in mental health care, particularly in non-clinical, lower-risk scenarios. Confidentiality and legal privilege are still of great concern. Here are some areas where AI can be helpful and possible considered safe.
1. Grounding Exercises and Stress Management
AI can be effective for providing structured exercises aimed at helping individuals manage stress and anxiety. Grounding techniques, such as mindfulness prompts, deep breathing exercises, or body scans, can be guided by AI to help users calm themselves during moments of distress. These exercises can be useful when someone needs immediate relief. Keep in mind that any prompts you feed it could be used to market to you, so it's best to keep them vague.
2. General Mental Health Information
AI can be a great resource for providing general information on mental health topics. This could include coping mechanisms, self-care tips, or overviews of various therapeutic modalities. It can serve as a starting point for individuals looking to educate themselves on mental health, but is not a substitute for tailored, personal guidance.
3. Cognitive Behavioral Techniques (CBT) for Mild Symptoms
For individuals dealing with mild anxiety or low-level stress, AI-driven programs can offer basic Cognitive Behavioral Therapy (CBT) tools. These programs can guide users through thought records, identify cognitive distortions, and help them develop healthier thought patterns. However, it’s important to remember that AI is not confidential and cannot provide personalized feedback. Those with more complex or serious conditions should seek human intervention.
4. Mood Tracking and Monitoring
If confidentiality is not a concern to you, AI can help you track moods, behaviors, and symptoms over time, providing insights into potential patterns. While AI is not equipped to interpret complex mental health conditions, tracking can be useful for those seeking to understand how their moods fluctuate. This data can be shared with a licensed therapist to inform their work.
5. Coping Strategies
AI can provide guidance on specific coping strategies. For example, AI can suggest action steps if someone feels overwhelmed or is experiencing a panic attack, like practicing deep breathing or reaching out to a support person. While AI can help users in these moments, it’s crucial to have a human therapist available for more personalized and ongoing support.
The Human Element in Therapy Cannot Be Replaced by AI
Alexis Harney, LMFT
Alexis is a licensed therapist working with adults and teens in California & Florida. She emphasizes a holistic, relational approach to achieve the best possible therapy outcomes.
Therapy is not just about exchanging information or guiding individuals through exercises. It is a dynamic, relational process that relies heavily on human emotions, connection, and the ability to navigate complex interpersonal dynamics. AI fundamentally lacks the capacity to engage in the rupture and repair process, or to examine transference and countertransference. These components of therapy are essential for long-term healing and personal growth. They cannot be replicated by an algorithm. As such, AI should only be viewed as a supplement to therapy, not a replacement for the deep, transformative relational work that human therapists provide.
Daniella Mohazab, AMFT
Daniella is a registered therapist offering trauma therapy in California. She uses a grounded, relational style to help process and synthesize human experience.
Until consumer protection regulations are put in place, and AI proves itself capable of performing all the essential tasks of a therapist, it’s best to treat it as a supplement.
Ready for Convenient, Personalized, Human Therapy?
While AI may offer convenience, it can never replace the depth and human connection that therapy provides. If you're seeking flexibility and accessibility, online therapy can offer much of the same convenience while ensuring you receive the confidential and nuanced care that only a licensed therapist can provide.
Connect with a real therapist from the comfort of your own home. Schedule a consultation today and experience the power of human connection.
