What is an AI Scribe?
Imagine a virtual assistant that listens to your conversation with your doctor and automatically writes up the clinical notes. That's essentially what an AI scribe does. Using natural language processing, these tools record, transcribe, and summarise
consultations, drafting everything from patient histories to referral letters. For doctors, who can spend hours on administrative tasks after a full day of seeing patients, the appeal is immense. The technology promises to free them from their keyboards, allowing for more direct, face-to-face interaction with patients. It’s a tool designed to tackle one of the biggest drivers of professional burnout: the overwhelming burden of documentation.
The Rush to Adopt
The uptake in Australia has been remarkably swift. A poll by the Royal Australian College of General Practitioners (RACGP) revealed that the use of AI scribes by doctors nearly doubled, rising from 22% in August 2024 to 40% by November 2025. This rapid adoption is fueled by the promise of efficiency and a return to the core of medical practice: patient care. Some AI providers claim their tools can increase a health professional's revenue by as much as 30% without adding hours, a claim the federal health department has noted has implications for Medicare costs. For many GPs, the technology is also a compliance tool, providing detailed records that can be useful in the event of a Medicare audit. Proponents argue it leads to better, more detailed notes and greater job satisfaction.
Privacy, Consent, and Accuracy Concerns
Despite the benefits, the rapid rollout has prompted serious warnings from the federal health department and patient advocates. A key issue is privacy. When a conversation is recorded, where does that data go? Some suppliers may be unaware that their platforms send patient data outside Australia, where privacy protections may be weaker. This has created what some experts call the 'Wild West' of medical technology. Then there is the question of consent. While many doctors are diligent about asking for permission, some patients have reported seeing posters in waiting rooms that imply consent is given simply by being there. Furthermore, the Australian Health Practitioner Regulation Agency (AHPRA) has stated that doctors can decline to see patients who do not consent to using an AI scribe, putting patients in a difficult position.
The Regulatory Lag
A significant part of the problem is a regulatory gap. Currently, most AI scribes are not classified as medical devices by the Therapeutic Goods Administration (TGA) as long as they don't offer diagnostic advice. This means they exist in a space with little direct oversight. The federal health department has acknowledged the technology has little oversight and that some companies market their products as being outside regulatory levers. This patchwork of regulation, split between bodies like the TGA, AHPRA, and the Office of the Australian Information Commissioner, has struggled to keep up. While the RACGP has issued guidelines, it stresses that the ultimate responsibility for the accuracy of notes and obtaining consent rests with the doctor.
The Dangers of an Inaccurate Record
Beyond privacy, the accuracy of AI-generated notes is a major concern. Like many large language models, AI scribes can make mistakes or even 'hallucinate'—inventing information that is false. Studies have found factual errors and significant omissions in AI-produced notes, such as incorrect diagnoses or medication lists. An error in a patient's medical record could have serious, even life-threatening, consequences. While doctors are expected to review and correct all AI-generated notes, the reality of a busy practice means errors could slip through. The RACGP has highlighted that these tools primarily produce free-text notes and often fail to populate structured data fields, which can lead to incomplete records.
