The promise of e-mental health chatbots is compelling: always on, stigma free and scalable support for populations chronically underserved by mental health systems. Tools like Woebot and Wysa attracted substantial investment and peer-reviewed attention with some RCTs suggesting modest reduction in depression anxiety symptoms for mild-to-moderate presentations (Hua et al., 2025; JMIR, 2025) and with Wysa even receiving FDA Breakthrough Device designation and CE-mark status in the EU this suggests genuine clinical credibility.
But who is using these tools, why, and on whose terms? In Aotearoa New Zealand psychological distress among 15–24-year-olds has tripled from 7.7% in 2014/15 to 22.9% in 2024/25 while access to mental health services for this age group has fallen 20% over 5 years(Mental Health and Wellbeing Commission, 2025). Pacific youth report distress rates of 38%, and research consistently shows the system is less responsive to the needs of Māori, Pacific and people with disability. Into this gap digital tools are being positioned as a solution: Whakarongorau Aotearoa, launched an AI concierge platform in 2025 as a “digital front door” to mental health services. The question is whether this fills a genuine care gap, or conveniently papers over a structural one. Meanwhile, the data these platforms collect deserves scrutiny: Intimate disclosures of trauma, distress or even suicidal ideation becomes training data shared between third parties under vague “service improvement” clauses. For communities already navigating mistrust of health systems the stakes of that data relationship are particularly high.
With the chatbot market estimated to reach $6.51 billion USD by 2032, deeper concerns remain unresolved: How safe are these tool? Who do they serve, or exclude?
From a patient-centred perspective, the key question may not be whether chatbots “work”, but for whom, in what contexts and as part of what care pathway and whether users meaningfully consent to what happens to their data along the way.