The Pros and Cons of AI in Mental Health

A Therapist's Journey Through AI Anxiety and Discovery (Part III)

Will AI make counsellors and therapists obsolete? Is AI a safe alternative to therapy? These are lively debates amongst mental health care providers, and ones that provoke some anxiety, not just around job security, but primarily around concerns for our clients' welfare. I, like so many people, are beginning to explore and question the role of AI in our real lives. In this post I explore the pros and cons of AI in mental health services, its potential and limitations.

Where AI Enhances Therapy

AI offers several benefits that can improve therapy for both clients and therapists. Some key advantages include:

• Increased Access to Care

AI-powered chatbots and apps provide mental health support anytime, anywhere. For people who cannot visit a therapist regularly due to location, cost, or stigma, AI tools offer a way to get help quickly. For example, apps like Woebot and Wysa use AI to guide users through cognitive behavioural techniques. That access to care may, in the interim, help someone feel heard. Because AI is available 24/7, it can offer psychoeducation, grounding exercises, or prompts to practise coping skills when it’s late at night or you’re between appointments.

• Offer Skills and Structure Between Sessions

AI can tailor exercises and reminders based on individual needs. This helps clients stay engaged between sessions. For example, an AI app might suggest mindfulness exercises when it detects rising stress levels from user input, or challenges unhelpful thoughts or track mood. Some research has shown short‑term symptom improvements for anxiety or depression. One randomised study of an AI “therabot” reported greater reductions in depression and anxiety over 4–8 weeks compared with a waitlist, with users rating the relationship with the bot as surprisingly collaborative. NEJM.

• Efficiency in Routine Tasks  

Therapists spend time on paperwork, scheduling, and note-taking. AI tools manage appointments and send reminders, reducing missed sessions. They can also prompt clients to complete homework or self-care tasks between visits. Some platforms use AI to transcribe sessions and highlight important points. This can free up more time for direct client interaction.

• Data-Driven Insights into Client Mental State

AI can analyze patterns in speech, writing, or behaviour to help therapists understand clients better and possibly predict where mental health issues are escalating. This can reveal hidden emotions or risks that might be missed in conversation. For instance, AI can detect changes in tone or word choice that suggest worsening depression.

• Virtual Reality Therapy  

AI combined with virtual reality creates immersive environments for exposure therapy. This helps clients face fears like phobias or PTSD triggers, or deal with addiction and craving in a controlled way.

Therapist taking notes with AI analysis on tablet

Where AI Falls Short in Therapy

Despite its benefits, AI also raises concerns that therapists and clients should consider:

• AI and Lack of Human Connection

Therapy relies heavily on empathy, trust, and understanding, something AI cannot actually feel. Some clients may feel isolated or misunderstood when interacting only with AI.

• Creating Unrealistic Expectations in Human Relationships

Though AI might be available 24/7, and it may seemingly give us what we want to hear in a tone we want to receive it, that's not what happens in real life. Humans have to navigate difficult emotions in relationship with others, as well as enjoy the positive ones. That isn't just in an intimate partnered relationship, but with friends, co-workers, family and with professionals, like therapists. We don't always get it right, and that ability to tolerate discomfort and step into difficult conversations is what makes real connection possible. If we stop doing so, do we stunt our interpersonal skills? Do we risk disappointment if others can't behave like or be our compliant AI partner?

• Risk of Over Reliance & Loss of Personal Resourcefulness

We might rely too much on AI tools and delay seeking professional help when needed. We may develop an unhealthy and/or time consuming habit of turning to AI instead of human support or engagement with a human world.

• Bias (including AI sycophancy)

AI systems learn from the data we give them. If most of the information AI uses to 'learn' comes from particular cultures, genders, religions, political viewpoints, then it can approach you with that lens. That lens may not be appropriate or culturally sensitive. for you. Moreover, AI collects our beliefs, bias, and concerns in the words we provide. It may reflect back to us the same bias, thereby reinforcing our thoughts, e.g. the tool we are relying upon becomes sycophantic in agreeing with us, instead of helping to gently challenge our way of thinking.

• Limited Understanding of Complex Issues  

AI works best with clear patterns and rules. It struggles with complex, nuanced human emotions and situations. For example, AI might not handle trauma or crisis situations as effectively as a trained therapist.

• Lack of Ethical Parameters  

AI works with data, not with a sense of ethical boundaries and professional responsibility. AI can make mistakes and may not recognise risk reliably. An unstructured, informal AI system without boundaries may operate in ways that hurt, rather than support clients. Professional bodies are being called upon to provide parameters for ethical use of AI. Ethical reviews emphasise strict testing and human oversight/governance for this reason Societies (MDPI)

• Privacy and Data Security

When you use AI tools you may be sharing sensitive personal information that cannot be withdrawn at a later date. There is a risk of data breaches or misuse. Even though therapists could use AI for session notes, is this information protected? Clients must be informed about how their data is stored and therapists must think carefully how the data is protected.

Looking Ahead: The Future of AI and Therapy

AI will continue to evolve and become more integrated into mental health care. My hope is that it will make the processes of providing support easier. For clients, this means more options to engage mental health tools.

The key is balance: using AI to improve access and efficiency while preserving the human connection that makes therapy effective. AI should complement, not replace, therapy with another person.

I heard a conference speaker recently state that AI Therapy doesn't actually exist. At the heart of therapy and counselling is a relationship, one in which the individuals have experiences of the relationship, of themselves, and the other. Though AI gives the illusion of interacting with us, it isn't capable of care, empathy, or authentic human connection.

I wonder if AI technology might cause us to reflect/re-examine the profound impact we feel if another, even a chatbot 'hears' us. How might the technology help us improve our listening skills? At the same time, I hope we can hold onto the desire to engage with other humans, to be challenged by and rewarded by the opportunity for interpersonal relationships.