37% of UK Adults Already Use AI for Mental Health — NHS Report Maps the Reality

The question is no longer whether AI will enter mental health practice. A third of UK adults are already using it — and the majority are choosing general-purpose chatbots, not purpose-built mental health tools. The NHS Confederation's new briefing maps this reality and begins to outline what responsible adoption looks like within clinical services.

The consumer reality

The most striking number: 66% of those using AI for mental health are using ChatGPT, Claude, or similar general-purpose tools. Not Woebot. Not Wysa. Not any of the purpose-built, clinically validated platforms. Patients are self-prescribing AI therapy from tools designed for writing emails and code.

This creates a governance gap that no amount of clinical AI regulation can close. The MHRA's forthcoming framework will set standards for purpose-built mental health AI — but it cannot regulate how a patient uses ChatGPT at 2 AM during a crisis. The briefing acknowledges this tension without pretending to resolve it.

What clinicians should know

The report distinguishes three tiers of AI in mental health: consumer self-use (unregulated, already widespread), clinical decision support (augmenting practitioner judgement), and patient-facing clinical tools (requiring MHRA certification). Each tier has different risk profiles, liability structures, and governance needs.

For NHS trusts already piloting AI tools, the briefing provides a practical framework: clinical safety assessments, data governance requirements, workforce training needs, and patient consent models. For private practitioners, the principles transfer even if the institutional structure does not.

The MHRA framework expected in 2026 will be the first UK-specific regulatory standard for AI in healthcare. Until it arrives, clinicians operate in a grey zone — and this report is the most practical guide to navigating it.