AI Therapy: Exploring the Future of Mental Health Support

Generation Z has adopted an AI-driven lifestyle, where AI chatbots dictate everything from breakfast recipes to assignment scripts. Nothing is off limits when dealing with Artificial Intelligence, and what that also means is you can now seek help from AI tools to help you manage life stresses, combat relationship challenges, process grief, and cope with loss. Cost-free and accessible with just a few clicks on a screen, AI is widely used for personal assistance, work, school, entertainment, research, and more. A 2023 survey of psychiatrists affiliated with the American Psychiatric Association (APA) revealed that 44% had used the 3.5 version of ChatGPT and 33% had used the 4.0 version to “‘assist with answering clinical questions.’” The research also found that 70% of psychiatrists surveyed indicated that they “somewhat agreed/agree” that with AI tools, documentation will be/is more efficient. m

Millions of people have been actively seeking mental health support from AI-powered mental health services. Therapy apps like Wysa and Youper have facilitated easy access to therapeutic treatment, especially in underserved communities. One can also adjust the AI system to adopt specific attributes or characteristics required during conversation. 

Many of the million mental health wellness app users have admitted that in times of depression and anxiety, moving out of their comfort zones and getting physical therapy seemed overwhelming. These apps, however, made therapy available to them in the convenience of their homes. The shift from human therapists to AI-driven therapy raises grave concerns regarding the authenticity of treatment and counseling. 

How does AI Therapy Work?

Artificial intelligence systems operate through user interaction interfaces like virtual agents, voice assistants, and chatbots, which determine how therapy takes place. These platforms are designed to understand human language through natural language processing (NLP). This technology reads into words, detects patterns, and traces the emotion behind the input to tailor a meaningful reply. It helps track and monitor behavioral patterns and underlying irrational beliefs through speech which determines the incorporated therapeutic technique to be utilized by the AI during the conversation, such as Cognitive Behavioural Therapy (CBT), Mindfulness-Based  Therapy, or Motivational Interviewing. 

Depending on how advanced the AI model is and the amount of data shared, it adapts to the conversational style and tone of the person and adjusts the preferred form of support being provided (motivational, empathetic, reflective). AI systems are trained to learn over time by tracking mood history and frequently occurring issues (health, family, anxiety). Furthermore, these apps can be used by the user to develop personalized interventions or exercises to cope with the issues. 

How Well do AI Bots Measure up to Traditional Psychological Therapy Standards?

According to the American Psychiatric Association (APA), effective psychological therapy encompasses the following characteristics:

1. Therapeutic alliance - trusting and collaborative relationship between therapist and client
2. Confidentiality and ethics - strict adherence to privacy rules and professionalism
3. Evidence-based practice - use of scientific approaches (CBT, DBT, RET)
4. Professional training and accountability - therapist must be licensed, supervised, and accountable to the code of ethics
5. Individualized care - treatment per the client's history and needs
6. Empathy and emotional attunement - the ability to deeply understand and connect with the client
7. Goal-oriented structure - sessions aimed toward therapeutic goals and relief
8. Crisis management and risk assessment - trained response to self-harm or suicide risk
9. Cultural sensitivity - respect and understanding of the client's background

Let's examine how AI therapy aligns with these key principles:

• Therapeutic alliance: absent - lacks human understanding and intuition
• Confidentiality and ethics: limited - data may be stored and shared in AI systems
• Evidence-based practice: limited - techniques lack flexibility and authenticity 
• Professional training and accountability: absent - AI tools do not meet formal standards of psychotherapy
• Individualized care: limited - AI may track mood trends but fails to provide the appropriate psychological-emotional cure
• Empathy and emotional attunement: absent - AI may be trained to mimic emotional language but it remains artificial
• Goal-oriented structure: present - AI may be useful in providing a structured program to meet goals 
• Crisis management and risk assessment: absent - AI cannot intervene in times of crisis
• Cultural sensitivity: limited - AI aims to be inclusive but it remains biased due to a lack of clear cultural understanding 

We can conclude that AI therapy is efficient in the role of providing goal-oriented interventions but falls short of most other standards required for the appropriate treatment of psychological issues. AI therapy continues to attract users due to its easy accessibility but fails to promote effective therapy resulting in a misunderstood idea of psychotherapy. Effective human-centered therapy includes core values as given by Carl Rogers, such as unconditional positive regard, empathy, and congruence. 

Is AI Always Fair?

In an informal test by a colleague, she pretended to be a victim of emotional abuse and received empathetic responses from a chatbot. But when she revealed that she was actually the abuser, the chatbot continued to offer emotional validation, failing to apply any moral or contextual judgment. This shows how AI systems may lack ethical filters and are easily manipulated depending on the persona the user presents.

AI tools are designed to agree with and support the user. In many cases, when faced with a situation, the AI system tends to favor the user's perspective, even if that support is not ethically appropriate. This tendency for AI to always align with the user's decisions encourages people to rely on these tools for validation. Many individuals prefer to be affirmed in their beliefs rather than seek out challenging truths. Consequently, AI often tells users what they want to hear instead of what they need to hear during crucial moments.

Decoding Biasness

1. Western-Centric Thinking

AI therapy tools are mostly trained on datasets dominated by white, English-speaking users. Which means they reflect the emotional norms and help-seeking styles of that specific cultural group. For instance,  someone from South Asia expresses mental distress not through “I feel sad,” but through “My chest feels heavy” or “There’s a constant weight on my head.” In many cultures, emotions are physical before they are verbal. Furthermore, Asian patients have reported that their physical symptoms were called to be 'exaggerations' by AI tools.

Therefore, AI can easily misread or even dismiss such expressions as irrelevant or illogical. This leads to responses that feel alien, even invalidating, especially to those whose mental health needs are already shaped by stigma and cultural complexity.

2. Gender and Identity Bias

Gender-diverse users - non-binary, transgender, and genderfluid users have reported that AI chatbots have misgendered them, ignored corrections, or responded with outdated scripts implying confusion or illness.

Some users have been met with responses that subtly frame their identity as a problem to be solved. This poses a huge block in societies still evolving and adjusting.  Therapy, whether human or digital, should be a space of affirmation. When AI misses that mark, especially with people who are already navigating marginalization, it becomes unsafe.

3. Life or Death Consequences

In 2024, a deeply concerning case emerged where an AI chatbot allegedly encouraged a suicidal teenager to act on their thoughts. This case was pivotal as it clarified that AI doesn’t always know when a user is in danger. Earlier tests showed similar results. A user types, “I want to disappear.” Instead of digging deeper or redirecting them to support, the AI replies, “Everyone needs a break sometimes.” This marks a danger for someone at their lowest. The gap here is more than technical. It’s ethical. Unlike trained professionals, AI doesn’t assess risk, doesn’t pick up urgency in your tone, and doesn’t escalate when it should. 

4. The Black Box Problem

Ever wonder how a chatbot comes up with its replies? Truth is, we don’t really know. AI systems work like black boxes, meaning they process your words through invisible layers of pattern recognition and statistical guesswork. But therapy isn’t guesswork. It’s full of moral, cultural, and emotional gray zones. If someone asks, “Should I leave my abusive partner?”, you don’t want an answer based on probability. You want insight, context, and ethical clarity. AI doesn’t have that compass. It responds with what’s most likely to be relevant, not necessarily what’s most responsible.

Hybrid Models: Future of AI Therapy

AI is undoubtedly powerful and can process what the human brain can’t, like patterns, trends, summaries, and even risk detection. The hybrid model, a possible and productive future of AI therapy, isn’t about AI replacing therapists but instead about AI assisting them. Like a quiet helper in the background. 

For instance, you talk to an AI chatbot mid-week when you’re spiraling at 2AM and it listens, comforts, and notes what triggered you. Before your actual therapy session, your human therapist gets a summary of what happened, what mood changes occurred, and what patterns are repeating. This way, the therapist can walk into the session already aware, picking up from where you stand at the moment without having to begin from scratch.

AI could be used to:

• Track emotional shifts over time – not through vague guesses, but actual data.

• Spot red flags – like rising mentions of hopelessness or isolation.

• Help you journal, reflect, or do CBT exercises when your therapist isn’t available

These AI uses are drawn from the fact that AI is most efficient at formulating goal-directed plans and tracking patterns, as previously noted. 

Conclusion 

Currently, while AI can be convenient, friendly, and agreeable, it often creates an illusion of intimacy. One should not expect intuitive understanding and connection from AI bots. Psychotherapy involves much more than just the words a client expresses; it also includes non-verbal cues such as a tear shed at a specific thought, a breaking voice during storytelling, hesitation, and slips of the tongue. These elements combined enable a therapist to truly grasp the human experience. Additionally, AI does not heal, it merely validates. Healing requires truth, patience, and emotional attunement, qualities that AI lacks. Therefore, it is crucial to differentiate between human therapists and AI therapy to promote mental wellness across populations.

***