top of page

Is Your AI Therapist Actually Helping You?

  • Writer: Sophie Cresswell
    Sophie Cresswell
  • Aug 1
  • 7 min read

A man and a woman sit opposite each other in low chairs while they talk.

A balanced look at what AI can and cannot offer your mental health


There's a question I've been sitting with lately, one that keeps coming up in conversation, in the research I read, and honestly, in my own reflections about what it means to help someone.


People are talking to AI about their mental health. Not occasionally. Not just out of curiosity. Some people are turning to AI chatbots every single day, sharing things they haven't told anyone else – about their fears, their grief, the thoughts that loop at 3am, about their fall out with a friend or their challenge at work. In some cases, they're choosing AI instead of a human therapist.


I don't think that's something to dismiss or panic about. I think it's worth understanding.

So let's look at it honestly: what does AI actually offer, where can it fall short, and what does that mean for you and me (and your friend who’s spilling the tea to Claude every day!).


Why People Are Turning to AI for Mental Health Support

Before we get into the debate, we need to ask: why is this happening?


The answer, I think, is less about AI being extraordinary and more about convenience and the system around mental health care having some very real gaps.


Seeing a psychologist in Australia often means waiting weeks or even months, and the cost can be daunting. The Australian Psychological Society recommends a standard fee of about $300, so most sessions fall in the $150–$300 range. Medicare rebates - usually $98 to $145 - can bring that down, but for many people, especially young people or those in regional areas, finances are still a genuine hurdle to getting help. 


For many people, the hardest part of therapy may not even be the financial cost. It's the vulnerability. Sitting with a real therapist means letting someone witness your inner world, and that can feel intimidating. People worry about being judged, about getting emotional, or about revealing something they’ve never said out loud. An AI doesn’t have facial expressions or body language. It doesn’t react, it doesn’t get uncomfortable, and it doesn’t need anything from you. That can make it feel easier to open up, even though it’s not a substitute for the support a trained therapist provides. 


And modern life is busy! I am guilty of using my commute to try and knock extra things off my ‘to do’ list and there is a definite appeal to the efficiency of using AI to work some feelings out rather than carving out the time for myself in the rest of my life. 

So we open our phones instead of making an appointment. It feels easier. And it is understandable. 


What AI Can Do Well

I should be open from the outset – I am an AI fan. It is not perfect, but it can offer some tools and efficiencies in my life that I love. And I don't think AI support is worthless. In some situations, it offers something genuinely useful.


It's available when nothing else is. At 2am on a Tuesday when your anxiety is loud and your therapist’s next appointment is two weeks away, being able to talk to ‘something’ is valuable. Research published in the Journal of Medical Internet Research has found that AI-based mental health tools can reduce symptoms of anxiety and depression, particularly mild to moderate symptoms, and particularly when human support is not accessible. That's valuable.


It removes the initial barrier. For some people, talking to an AI first is a steppingstone. It can give them language for what they're feeling, help them rehearse saying things out loud before they're ready to say them to a person. A 2023 study from Stanford found that users of AI mental health tools reported feeling more prepared to seek professional care after using them. So sometimes the AI isn't the destination, it's a step on the journey.


It can provide psychoeducation consistently and patiently. AI is well-suited to explaining what anxiety is, walking someone through a grounding technique, or offering information about sleep hygiene without judgment and without fatigue. 


It meets people where they are. Not everyone is in crisis. Some people just want to process a hard week. They want to feel heard in a low-stakes way. AI can hold that kind of conversation without the overhead of scheduling, cost, or the emotional labour of showing up to a formal session.


Where AI Genuinely Cannot Go

And now for the part that matters just as much.


There is something that happens in a good therapeutic relationship that I don't believe can be replicated by an algorithm, no matter how sophisticated. Researchers call it the therapeutic alliance – the bond between therapist and client, built on trust, attunement, and genuine human presence. It is, consistently, one of the strongest predictors of therapeutic outcomes. It is also something that requires a person.


When I sit with someone, something passes between us that isn't just words. I notice the way their shoulders drop when they finally say the thing they've been holding or how their breathing changes when tension is released. I feel the shift between us when someone moves from defense to openness. That's not a metaphor. That is relational data that shapes how I respond, what I reflect back, when I stay quiet. An AI doesn't have that. It processes language. It cannot read the room it isn't in.


When two people meet (in person or virtually), your body naturally responds to theirs. Your breathing settles, your heart rate slows and you start to feel a little more grounded simply because you are connecting to another calm nervous system. This is co-regulation and it helps you stay steady enough to talk about things that might otherwise feel overwhelming. A therapist can notice when you tense up or when you soften and they can adjust in the moment so you feel supported and understood. That kind of real-time, human-to-human calming is something language processing cannot offer because it depends on two people sharing the same space. 


There are also clinical limits that matter enormously. AI is not equipped to assess suicide risk. It is not equipped to identify when someone's symptoms suggest something more serious, like a psychotic episode, a dissociative process, the early signs of a manic phase. It cannot make a differential assessment. And in some cases, the consequences of missing those things are not recoverable.


One of the things that concerns me most is the risk of what I think of as therapeutic mimicry. AI can sound warm and validating enough that someone feels understood for a moment, but real validation is not just about kind words. It comes from another person taking you in, noticing your emotional cues and responding in a way that fits what you are actually feeling. That kind of attuned, human validation helps people feel safe enough to change. When the support is only simulated, it can create the feeling of being heard without the deeper shifts that come from genuine connection. And for some people, that illusion of support can delay them from reaching out for the real help they need. 


There is also the question of what a person takes from the interaction. Therapy, at its best, is relational learning. It changes how you are with people, not just how you feel in the moment. A conversation with an AI does not offer that. It ends when you close the app.


The Question of Dependency

The other elephant in the room is dependency. Some people are developing what feels like an attachment to AI tools. They find it easier to talk to an AI than to the people in their lives. They prefer the consistency, the lack of judgment, the sense of being understood without risk. 


Honestly, I understand the appeal. But it is worth pausing to ask what we are moving toward. If the aim of good mental health support is to help someone live more fully in their own life with more connection, more capacity and more presence, then relying heavily on AI may not take us in that direction. 


Western culture often teaches us that healing is a solo project, something we can think our way through if we just try hard enough. Yet most of the research on recovery and wellbeing points to the opposite. We heal in connection. We grow through relationships that challenge us, soothe us and help us see ourselves more clearly. Human relationships can be messy and uncomfortable, but they are also the place where real change becomes possible. 


A Thoughtful Way Forward

I'm not arguing that AI has no place in mental health. I think it can play a meaningful role, particularly as a first point of contact, a support between sessions, a way of reducing barriers for people who would otherwise have nothing.


But I do think it's worth being honest about what it is and what it isn't.


If you've been using AI to help you process things, that's okay. If it's helped you find words for what you're feeling, or get through a hard night, that’s great. And if part of you suspects that what you're carrying might benefit from more than an app can offer: that instinct is probably worth following.


Real support exists. It doesn't have to cost what you think it costs. It doesn't have to wait as long as you think it does.


You deserve the kind of help that actually knows and cares that you're there.


If this resonated with you, Redefine Life offers a free introductory call and low-cost counselling sessions available online, Australia-wide. Sometimes the first conversation is the hardest one to have, and it's okay to start small.



References

  • Fitzpatrick, K. K., Darcy, A., & Vierhile, M. (2017). Delivering cognitive behavior therapy to young adults with symptoms of depression and anxiety using a fully automated conversational agent (Woebot): A randomized controlled trial. JMIR Mental Health, 4(2), e19.

  • Norcross, J. C., & Lambert, M. J. (2018). Psychotherapy relationships that work III. Psychotherapy, 55(4), 303–315.

  • Stade, E. C., Lourie, C., & Woebot Health (2023). Artificial intelligence in psychotherapy: Exploring user perceptions and readiness for human care. Stanford Social Innovation Review.

  • Vaidyam, A. N., Wisniewski, H., Halamka, J. D., Kashavan, M. S., & Torous, J. B. (2019). Chatbots and conversational agents in mental health: A review of the psychiatric landscape. The Canadian Journal of Psychiatry, 64(7), 456–464.

  • Woebot Health (2021). The state of mental health in America: How AI-powered mental health tools are filling gaps in care. Woebot Health White Paper.



Written By: Sophie Cresswell

Reviewed By: Chantelle Gagachis

Comments


bottom of page