A person wakes at 3 a.m., overwhelmed, and opens ChatGPT instead of calling a friend or booking a therapy appointment. This scenario is no longer unusual. Around the world, AI chatbots are functioning as informal companions, coaches, and, for a growing number of people, unofficial therapists. Understanding what that actually means, and where it leads, requires looking past the headlines about AI disrupting healthcare and asking a more precise question: what can this technology genuinely do for mental health, and where does it fall short?
Why People Turn to Chatbots for Emotional Support
The appeal of AI as a mental health resource is not hard to explain. Chatbots do not judge. They are available at any hour. And unlike mental health services in countries such as New Zealand and Australia, they do not place people on lengthy waiting lists. For someone in distress who cannot access professional care quickly, an AI that responds with apparent empathy can feel like a meaningful alternative.
Research has found that many users already turn to AI to discuss personal struggles, seek emotional support, and reflect on their mental state. Some studies suggest that carefully designed AI tools can help reduce symptoms of anxiety and depression when used appropriately. AI is also showing early promise in helping people practise cognitive reframing, which involves learning to interpret difficult situations in alternative, less harmful ways.
These are real capabilities. They are not trivial. But they exist alongside serious limitations that researchers, clinicians, and regulators have been raising with increasing urgency.
The Gap Between Sounding Empathetic and Being Clinically Safe
Here is what most coverage of AI in mental health tends to underplay. An AI system can sound understanding without actually understanding the person behind the conversation. That distinction matters enormously in a clinical context.
AI systems can generate inaccurate advice. They can agree with or reinforce harmful beliefs rather than redirecting a person toward appropriate help. They can miss signs of crisis. And unlike mental health professionals, they are not held to the same professional or regulatory standards when something goes wrong. Mental health care depends on trust, clinical judgement, and human connection in ways that go well beyond information delivery.
Recent research also points to a subtler risk: over-reliance. Because AI tends to respond in ways that feel supportive and validating, users may accept its guidance without questioning it or seeking professional help. In mental health settings, that uncritical trust can have serious consequences.
This is why many experts frame AI as a tool to support mental health care rather than something that can or should replace it. The framing is not diplomatic hedging. It reflects a genuine understanding of what the technology is and is not capable of.
Where AI Might Actually Make a Difference: Earlier Detection
One of the more concrete and promising applications comes from research being conducted at the University of Auckland’s 2DN research group. The team is investigating whether AI can detect signs of depression earlier by analysing patterns in how people communicate.
Depression affects communication in measurable ways. Changes in speaking rate, pauses, tone of voice, word choice, and emotional expression can all provide clues about a person’s mental state. These are examples of what researchers call “digital biomarkers,” measurable patterns in behaviour or physiology that can signal something about health. Researchers are also exploring other potential biomarkers, including facial expressions, sleep patterns, and physical activity.
The goal of this work is not to have AI diagnose people or replace clinicians. The aim is to develop screening and monitoring tools that can flag individuals who may benefit from further professional assessment. The analogy the researchers use is instructive: a wearable device that detects unusual heart activity does not replace a cardiologist. It gives the cardiologist another piece of information to work with.
That framing captures something important about where AI adds genuine value in healthcare. It handles pattern recognition at a scale and consistency that human observation cannot match. It does not replace the clinical judgement that follows.
What This Means Beyond the Clinic
The broader implications extend in two directions. One is about access. AI has the potential to reach underserved communities, reduce barriers to seeking help, and personalise support based on individual needs when sufficient high-quality data are available. For populations where mental health services are scarce or stigmatised, that is not a minor benefit.
The other direction involves risk. Mental health data is among the most sensitive information a person can share. Privacy, security, and informed consent require careful protection. AI systems can also inherit biases from their training data, which means they may work less effectively for certain populations, precisely the groups that often face the greatest barriers to care.
The technology’s role in mental health will grow. That much seems clear. The question is whether it grows in ways that genuinely support people or in ways that create new vulnerabilities while appearing to help.
In Short
AI chatbots are already functioning as informal mental health resources for many people, and research suggests they can provide real support in specific, well-defined contexts. But they cannot replicate clinical judgement, human empathy, or professional accountability. The most credible path forward treats AI as an augmentation layer: a tool that helps people understand their mental wellbeing earlier and gives clinicians better information to act on. Technology recognises patterns. People provide the trust, empathy, and judgement that mental health care actually requires.
Based on reporting from The Conversation Technology.