Over the past year, it has become almost impossible to ignore the rapid rise of artificial intelligence in everyday life. From writing emails to answering complex questions, tools like ChatGPT and other AI platforms are quickly becoming part of how people think, work, and solve problems.
It is no surprise that this shift is beginning to reach the world of mental health.
More and more, clients are experimenting with AI in ways that intersect with therapy. Some are using it to process thoughts, generate coping strategies, rehearse difficult conversations, or even simulate emotional support. At the same time, therapists are beginning to explore how AI might support their own work, from administrative tasks to clinical reflection.
But as with any major shift, the question is not just what AI can do. It is also what it should do, and where the boundaries need to be.
What AI Is Already Doing in Mental Health
AI is not a future concept in psychotherapy. It is already here.
Some of the current use cases include:
- Chatbots designed to provide mental health support, including CBT-based tools
- Mood tracking and journaling platforms with AI feedback
- Psychoeducation and coping strategy generation
- Therapist support tools such as note-taking, treatment planning assistance, and supervision reflection
For many people, these tools offer something traditional therapy does not always provide: immediacy, accessibility, and lower cost. Someone can open an app or browser at any time and receive a response within seconds.
Research on digital mental health tools suggests that AI-based interventions can increase access to support and provide short-term symptom relief, particularly for mild to moderate concerns (Miner et al., 2020). Barriers to mental health care, including cost, stigma, and availability, are very real. At the same time, AI operates in fundamentally different ways than human therapy.
What AI Cannot Replicate
At its core, psychotherapy is not just about information. It is about relationship.
Approaches like Emotionally Focused Therapy, attachment-based work, and experiential therapies emphasize that healing happens not only through insight, but through felt emotional experience within a safe relational context (Elliott et al., 2018).
AI can simulate empathy. It can generate validating language. But it does not experience emotion, nor can it truly attune in the way a human nervous system can.
There is no genuine co-regulation. No lived relational presence. No shared emotional field.
For many clients, especially those working through trauma, attachment wounds, or long-standing relational patterns, those elements are not optional. They are central to change.
The CRPO Perspective: Ethical and Professional Responsibility
In Ontario, the College of Registered Psychotherapists of Ontario (CRPO) has begun addressing the use of digital tools, including AI, within psychotherapy practice.
While guidance continues to evolve, several core principles remain clear:
- Confidentiality and privacy must always be maintained
- Therapists are responsible for understanding the tools they use
- Client consent is essential when integrating technology into care
- Technology should not compromise the quality or safety of services
These expectations are grounded in broader professional standards emphasizing ethical use of technology and protection of client information (CRPO, 2023).
In other words, AI is not prohibited, but it must be used thoughtfully, ethically, and within the therapist’s scope of competence.
This is particularly important because AI tools are not neutral. They are trained on large datasets, may contain biases, and do not always produce accurate or clinically appropriate responses (World Health Organization [WHO], 2021).
A Critical Concern: Client Privacy
One of the most important and often overlooked considerations in the use of AI in mental health is privacy.
Many AI tools process user input through external servers. This means that anything entered, including deeply personal or sensitive information, may be:
- Stored
- Reviewed for model improvement
- Subject to data breaches or security vulnerabilities
For clients, this raises important questions:
- Who has access to this information?
- Is it protected under health privacy laws?
- How is it being used or stored?
For therapists, the stakes are even higher. Entering identifiable client information into AI tools without proper safeguards may violate professional standards and privacy legislation (CRPO, 2023). Even anonymized information carries risk if it can be indirectly linked back to a client.
Because of this, regulatory bodies and global health organizations emphasize caution, transparency, and informed consent when integrating AI into care (WHO, 2021).
What This Could Mean for the Future of Psychotherapy
AI is not going away. If anything, its role in mental health will continue to expand.
We may see:
- Increased use of AI as a supplemental support tool between therapy sessions
- Greater integration into therapist workflows, including documentation and resource generation
- More conversations about digital ethics and competence in therapist training programs
- A widening gap between accessible AI-based support and relational, human-centered therapy
Rather than replacing therapists, AI may push the field to become clearer about what makes psychotherapy unique.
Not just advice. Not just insight. But relationship, attunement, and emotional presence (Elliott et al., 2018).
Holding Both Curiosity and Caution
As therapists, and as a profession, we are in a moment of transition.
There is value in being curious about what AI can offer:
- Increased access to support
- Tools for reflection and skill-building
- Opportunities to enhance aspects of care
But there is also a responsibility to remain grounded in what we know about healing:
- Relationship matters
- Emotional safety matters
- Privacy and trust are foundational
AI may change how people engage with mental health support. But it does not change what human beings fundamentally need when they are struggling. And that is something worth holding onto as the field continues to evolve.
