(Insights)

Using AI safely in a medical practice: what Ahpra expects.

Key points

  • Ahpra supports the safe use of AI, but practitioners remain responsible for the care they provide and must apply human judgement to anything AI produces.
  • Patients should be told when AI is used, and informed consent is generally needed when an AI tool uses their personal information, including before a consultation is recorded.
  • The evidence shows real but modest time savings: about 16 minutes less documentation per eight hours of patient care with ambient AI scribes.
  • A simple checklist covering data, consent, human review and escalation keeps adoption safe.

Why practices are looking at AI

Admin is wearing general practice down. Nearly seven in ten Australian GPs report burnout, and most are dissatisfied with their administrative burden.1 In 2024, 70% of GPs said they were concerned about their admin workload, up from 60% a year earlier.2 AI receptionists, ambient scribes and automation promise to take some of that load.

What Ahpra’s guidance says

In 2024 Ahpra and the National Boards published guidance on meeting professional obligations when using AI in healthcare.3 It doesn’t ban AI tools. It explains how existing obligations apply to them:

  • You stay accountable. Practitioners remain responsible for delivering safe, quality care whatever technology they use, and must apply human judgement to any output from AI.
  • Check what AI writes. The practitioner is responsible for checking the accuracy and relevance of records created using generative AI, including notes drafted by AI scribes.
  • Get informed consent. Where an AI tool uses a patient’s personal information, involve the patient in the decision and ideally record their response in the health record. Recording a consultation needs consent before recording starts.
  • Be transparent. Tell patients when AI is being used. How much detail is needed depends on the tool: one that records consultations calls for a fuller explanation than, say, image analysis software.
  • Approval isn’t a free pass. Even where a tool is approved by the Therapeutic Goods Administration, practitioners still need to understand it and check it is fit for purpose before clinical use.

State and territory laws on recording conversations can also apply, so check the rules where you practise.

What the evidence says about time savings

The largest study so far, published in JAMA in 2026, followed 8,581 clinicians across five US academic health systems.4 Using an ambient AI scribe was linked to about 16 fewer minutes of documentation per eight hours of scheduled patient care, with larger savings for clinicians who used the scribe consistently. Importantly, the study measured time, not documentation quality or patient safety.

Simpler automation can pay off too. A Cochrane review found SMS appointment reminders lifted attendance from 68% to 77%, at a lower cost per attendance than phone-call reminders.5

The lesson: AI and automation can genuinely give time back, but the gains are usually steady rather than dramatic, and they depend on good setup and consistent use.

A practical checklist for adopting AI safely

  1. Start with a problem, not a product. Define the task you want to improve and how you will measure it.
  2. Map the data. Know what patient information the tool touches, where it is stored and who can access it. Health information is sensitive information under Australian privacy law.6
  3. Write the consent wording. Decide how patients will be told about the tool and how consent is recorded, especially before any recording.
  4. Keep a human in the loop. Anything clinical, from notes to triage, is checked by a practitioner.
  5. Set escalation rules. An AI receptionist must hand urgent or clinical calls to a person and direct emergencies to 000.
  6. Check the vendor. Ask about security, data handling and where data is stored before you sign. Staying vendor-agnostic lets you choose on merit.
  7. Train the team. Make sure everyone knows what the tool does, its limits, and what to do when it gets something wrong.
  8. Review and adjust. Track errors, complaints and time saved, and be willing to switch off what isn’t working.

Want help applying this to your practice? See our AI and digital support.

  1. Royal Australian College of General Practitioners. General Practice: Health of the Nation 2025.
  2. RACGP Health of the Nation 2024 survey, early findings. newsGP, 2024.
  3. Ahpra. Meeting your professional obligations when using Artificial Intelligence in healthcare, 2024.
  4. Rotenstein L, et al. Changes in clinician time expenditure and visit quantity with adoption of AI-powered scribes: a multisite study. JAMA, 2026.
  5. Gurol-Urganci I, de Jongh T, Vodopivec-Jamsek V, et al. Mobile phone messaging reminders for attendance at healthcare appointments. Cochrane Database of Systematic Reviews, 2013.
  6. Office of the Australian Information Commissioner. Australian Privacy Principles guidelines, Chapter B: Key concepts.

(Next step)

Bring AI into your practice, safely.