The Consent Illusion: How One Melbourne Practice Exposes a Fault Line in AI Healthcare Adoption
A Melbourne psychiatrist is refusing to accept new patients who decline to have their sessions documented using artificial intelligence, according to a registration form reviewed by this publication. The form, distributed to prospective patients in May 2026, states that individuals who object to AI note-taking must ask their referring doctor to find an alternative provider. The policy has reignited a fraught debate about consent, autonomy, and the pace at which AI systems are being embedded into clinical practice.
The case crystallises a tension that healthcare systems worldwide are only beginning to grapple with: as AI documentation tools promise efficiency gains and reduced administrative burden, the question of whether patients can meaningfully withhold consent has become increasingly urgent. A referral requirement that effectively conditions access to a specialist on acceptance of algorithmic record-keeping is, for privacy advocates, coercive in all but name.
The Australian Medical Association's guidelines on AI in clinical practice acknowledge that patient consent must be informed and voluntary. What the guidelines do not specify is what constitutes a genuine alternative when a practitioner simply declines to offer a non-AI pathway. Critics argue this vacuum allows individual practices to set terms that circumvent the spirit of consent requirements, particularly in specialties where waitlists are long and options limited.
The healthcare sector has watched AI documentation tools proliferate with cautious optimism. Vendors market these systems as a solution to the administrative overload that contributes to clinician burnout—a problem documented extensively in Australian medical literature. Proponents argue that freed from the chore of note-taking, practitioners can spend more time with patients. The efficiency argument is not trivial: one widely cited study suggested Australian doctors spend nearly two hours on administration for every hour of direct patient contact.
Yet the consent question resists easy dismissal. Medical records carry sensitivities that go beyond ordinary data. Psychiatric notes in particular can contain material that patients share only because they trust the confidentiality of the clinical relationship. The introduction of a third-party AI system into that dyad changes its character, regardless of what privacy policies the vendor provides. Patients are being asked to trust not only their clinician but also a technology company, its data-handling practices, and whatever future uses AI-generated notes might be put to as training data or in automated decision-making.
The regulatory landscape in Australia currently lacks the specificity needed to resolve these tensions. While the Privacy Act establishes baseline obligations for health information, it was not drafted with AI documentation in mind. The Therapeutic Goods Administration oversees medical devices but has moved cautiously on software that sits at the boundary between tool and decision-maker. The Australian Information Commissioner can act on individual complaints but has not issued guidance specific to AI note-taking in private practice. Practitioners are, in effect, operating in a grey zone, setting their own terms.
The structural pattern here extends beyond this one practice. Across industries, the deployment of AI systems has frequently preceded the regulatory frameworks needed to govern them responsibly. Healthcare is not immune; if anything, the stakes are higher given the intimacy of the clinical relationship and the sensitivity of health data. What is notable about the Melbourne case is its frankness: rather than burying the AI requirement in fine print, the practice has made it a condition of access. That transparency may be preferable to the more common approach of introducing such systems without explicit patient awareness, but it does not resolve the underlying problem that saying no can mean going without.
Whether this represents a one-off policy decision by a single practitioner or a sign of broader normalisation remains to be seen. What the episode makes clear is that the question of how AI enters medicine cannot be left to individual clinicians and their vendors. Professional bodies, regulators, and patient advocates will need to develop clearer standards—not just for what consent looks like on paper, but for what genuine choice means when the practical alternatives are thin on the ground.
This publication has not verified the identity of the specific psychiatrist or practice named in the registration form.