You’ve probably already done it. Typed something into ChatGPT that you haven’t said out loud to anyone. A fear, a worry, something that felt too raw or too embarrassing to bring to a real conversation. And got back a response that felt surprisingly thoughtful.

That moment – of being “heard” by an algorithm – is something millions of people are describing, including right here in Australia. And it raises a question worth taking seriously: is using an AI therapist a genuine option, or does it carry risks that aren’t obvious in the moment?

In this post, we look at what the research actually says: what AI can do, where it falls short, and what it cannot replace.

Why Australians Are Turning to AI for Mental Health Support

The numbers are hard to ignore. According to a 2025 study using a nationally representative Australian community sample, nearly 1 in 10 Australians had asked ChatGPT a health question in the previous six months. A broader report on generative AI use put the figure for mental health-specific use at over 10%.

The reasons people give are understandable: availability, cost, and privacy. A psychologist costs money and requires an appointment. ChatGPT, however, is free and available at 2am when you can’t sleep. For many people, especially those who have been waiting weeks to see a mental health professional, or who aren’t sure whether what they’re experiencing is “bad enough” to warrant help, AI feels like a logical first step.

That empathy is important. There is no shame in reaching for what’s accessible when you’re struggling. But understanding the limits of AI therapy in Australia is also essential, because those limits matter clinically, and in some cases they matter urgently.

What AI Can Genuinely Do Well

To give an honest answer, AI does have real strengths in the mental health space, particularly when used as a complement to care, not a replacement for it.

AI tools can be helpful for:

  • Psychoeducation: learning about conditions like anxiety, depression, or OCD in plain language, at your own pace, without feeling put on the spot
  • Practising CBT skills: thought records, cognitive restructuring, behavioural activation. These structured exercises translate reasonably well to a chat format
  • Mood tracking: keeping a simple log of how you’re feeling between sessions, which can be genuinely useful data when you do see a psychologist
  • Initial reflection: articulating what you’re experiencing before a first session, which can reduce the anxiety of not knowing what to say

Where an AI therapist has some merit is in accessibility and immediacy. For someone who is new to therapy, uncertain about what they need, or waiting for an appointment, using AI to get some initial understanding of their experience is not inherently harmful, provided the situation is not acute.

The key phrase there is: provided the situation is not acute.

ai therapist

Where AI Therapists Fall Short: What the Research Shows

The clinical picture changes significantly when we look at research into how AI actually performs in higher-stakes mental health situations.

A March 2026 study from Brown University evaluated generative AI models, including ChatGPT, for therapy-style interactions and identified 15 distinct ethical risks. These included mishandling crisis situations, reinforcing harmful beliefs, and offering what researchers called “deceptive empathy”: responses that feel caring but lack genuine clinical understanding of the person’s situation.

In the same period, a survey by the American Psychological Association found that 94% of psychologists believed AI chatbots cannot treat mental health conditions with the appropriate level of nuance. The same data showed that AI responded inappropriately to mental health symptoms in at least 20% of interactions.

A 2025 Stanford University study tested five commonly used chatbots against simulated mental health crisis scenarios, including expressions of suicidal ideation, psychotic symptoms, and mania. None of the chatbots met clinical standards for crisis response. Some validated delusional thinking. Some encouraged behaviours that would be clinically contraindicated.

Closer to home, the Australian Psychological Society (APS) issued a public warning in 2026 that “Australians are at risk as AI and digital mental health services roll out without appropriate evaluation.” The APS position is clear: an AI therapist cannot provide clinical, individualised psychological care, no matter how sophisticated the technology becomes.

Consequently, this is not a minor caveat. It is a structural limitation.

The Problem No Algorithm Can Solve: The Therapeutic Relationship

Here is what often gets missed in the AI-versus-therapy conversation: the most consistent predictor of therapy working is not the type of therapy. It is the relationship between the client and the psychologist.

Decades of research into what makes therapy effective, across different conditions, different modalities, and different populations, point to the same finding. A strong therapeutic alliance, defined as a genuine, trusting working relationship between therapist and client, consistently predicts better outcomes. Moreover, a landmark series of meta-analyses found this relationship accounts for a significant proportion of therapeutic change, independent of the technique used.

What does this mean in practice? It means that feeling genuinely understood by another person, someone who holds your full history, who notices what you don’t say, who adjusts their approach based on how you’re doing that week, is not incidental to the process. It is the process.

An AI therapist can simulate the language of empathy. However, it cannot build a relationship. Unlike a human clinician, it does not know you across time. It cannot notice that your tone has shifted, that you seem more flat than usual, that something you mentioned three sessions ago is quietly connected to what you’re describing today. And it cannot be accountable for you, or to you.

What Human Therapy Actually Offers

At MyLife Psychologists, our first step when someone reaches out is always a careful intake process. We want to understand not just the presenting concern, but the person’s history, their way of relating to the world, and what kind of support is likely to actually help them.

Then we match. The psychologist you see is not randomly assigned. It is a considered decision, because we know that the right fit makes a clinical difference.

From there, the work involves evidence-based approaches tailored to your situation. Depending on what you are dealing with, this might include:

As a result, these approaches work. They are backed by decades of research. And they are delivered by a person who knows you, who is accountable to you, and who adapts their approach as you change.

That is not something an AI therapist, in Australia or anywhere, can currently provide.

ai therapy

Should You Stop Using AI Altogether?

Not necessarily. If you find it helpful for journaling, for learning, for practising a skill between sessions, or for getting your thoughts organised before a first appointment, that can be a reasonable use of the technology.

What we would caution against is using an AI therapist as a primary source of mental health support, particularly if:

  • You are dealing with something serious, complex, or longstanding
  • Your symptoms are getting worse, not better
  • You are in any kind of crisis
  • You have been relying on AI for weeks or months without accessing professional support

The question to hold is this: is this helping me move forward, or is it helping me feel managed without actually getting better?

Taking the Next Step

If you have been sitting with something for a while, relying on apps, podcasts, or AI conversations to get through, and part of you knows that what you are dealing with goes deeper than that, it may be time to speak with someone.

At MyLife Psychologists, you don’t have to have it all figured out before you reach out. Our Care Coordinator will take the time to understand what you are experiencing and help you find the right psychologist for you, from the very first call. Book a free 15-minute call with our Care Coordinator to find out more.

References

  • Ayers, J. et al. (2026). ChatGPT as a therapist? New study reveals serious ethical risks. Brown University/ScienceDaily.
  • American Psychological Association. (2026). Patients are bringing AI to therapy: APA Chatbots and Mental Health Survey.
  • Stanford University. (2025). New study warns of risks in AI mental health tools. Stanford Report.
  • Australian Psychological Society. (2026). Australians at risk as AI and digital mental health services roll out without appropriate evaluation.
  • Prevalence of ChatGPT use in Australia: nationally representative community sample. (2025). Digital Health Week.
  • Flückiger, C. et al. (2018). The alliance in adult psychotherapy: A meta-analytic synthesis. Psychotherapy, 55(4), 316-340.

Useful resources

  • Beyond Blue – Mental health information and support
  • Lifeline – 24/7 crisis support: 13 11 14
  • Headspace – Mental health support for young people aged 12–25
  • SANE Australia – Support for people affected by complex mental health issues
  • Head to Health – Australian Government digital mental health service finder

 

Disclaimer: This article is for general informational purposes only and is not a substitute for individual psychological advice, assessment, or treatment. Reading this content does not establish a therapeutic relationship. If you have concerns about your mental health, please seek support from a registered health professional.