AI mental health support is no longer a future problem. People are already using chatbots to talk about grief, panic, loneliness, relationships and distress, often at the very moment they need someone to respond.
That makes it worth taking seriously. Not because AI is either wonderful or terrible, but because it is available, persuasive, and increasingly part of how people try to make sense of their inner lives.
AI mental health support is already here
On The Nutters Club recently, Hamish Williams and I talked about the growing use of tools such as ChatGPT and Google Gemini for emotional support. The programme also tried it live, with Hamish talking to a chatbot about his disappointment after the Warriors’ finals loss.
The response was, in some ways, impressive. It acknowledged the feeling, asked a few reasonable questions, and later suggested practical ways to wind down. It sounded familiar because, at a surface level, a lot of supportive conversation does involve listening, reflecting and gently asking what might help next.
That accessibility matters. Counselling has waiting lists. Friends are asleep at 1am. Phone services may feel too difficult to call. A chatbot, by contrast, is immediate, private and endlessly patient. It is not surprising that some people find that useful.
Emerging research discussed on the programme suggests this is not rare. One Canadian study of about 40,000 students found roughly one in five had used AI for help with feelings or relationship advice. Whether we like it or not, this is already part of the mental health landscape.
What AI can and cannot do
The useful question is not, “Can AI help?” It clearly can help some people in some moments. A chatbot may be able to suggest a breathing exercise during panic, offer ideas for winding down, or help someone put words around what they are feeling.
But feeling supported is not the same as receiving mental health treatment.
Therapy is not only advice, information or soothing words. It is a relationship in which another person notices patterns over time, responds to what is not being said, carries concern, tolerates complexity, and can be accountable for the care they offer. A therapist can also make clinical judgements about risk, trauma, depression, anxiety and the many ways these can overlap.
AI can imitate parts of a caring conversation. It cannot actually care. That distinction may sound philosophical, but it becomes very practical when someone is severely distressed, suicidal, frightened of their own thoughts, or isolated from other people.
In those moments, AI should not be the destination. At best, it may be a bridge: something that helps someone hold on long enough to contact a real person, a crisis service, a GP, a counsellor, a friend or whānau.
Why human connection still matters
We have been here before, in a way. New technologies often arrive with a mix of excitement and panic. I have written previously about the tendency to ask whether a new platform is simply “bad for us”, when the more useful question is usually how, when and why we use it: Is Facebook bad for you?
AI raises a sharper version of that question because it does not just connect us to other people. It can mimic being another person.
That mimicry can be comforting. It can also be confusing. Human empathy is not only the right words in the right order. It is tone, timing, facial expression, shared vulnerability, and the bodily experience of being with someone who is affected by us. I have written elsewhere about how empathy is wired into human connection, and that embodied quality is not a minor detail.
None of this means people are foolish for using AI. If someone feels alone and a chatbot helps them calm down, that matters. But it is worth asking: does this help me move towards people, or away from them? Does it give me enough steadiness to seek support, or does it become a substitute for being known?
Listen to the episode
The full Nutters Club discussion explores both the promise and the discomfort of AI mental health support, including listeners’ very different experiences of using it. If you are curious, sceptical, or already using AI in this way, it is well worth listening to the whole conversation.
