AHPRA AI Guidelines Explained: What Australian Clinicians Actually Need to Do [2026]
AHPRA has published guidance on AI in healthcare. Most clinicians haven't read it. Here are the 4 obligations every AU clinician must meet when using AI in patient care — and what each one looks like in practice.
Quick answer
The AHPRA AI guidelines boil down to four obligations:
- Patient consent — verbal, documented, mandatory.
- Clinician responsibility — you read, edit, and sign the final note.
- Data location — offshore processing is a notifiable disclosure without explicit consent.
- Documentation integrity — notes must reflect the specific clinical encounter, not templated AI phrasing.
If your AI workflow meets all four, you're aligned with AHPRA. The rest of this article unpacks each one with worked examples.
Obligation 1: Patient consent — verbal, documented, mandatory
AHPRA AI guidelines treat patient consent as non-negotiable. The guidance is explicit: any use of AI in patient care that involves processing of personal information requires the patient's knowledge and consent.
The minimum standard:
- Verbal consent obtained at the start of the consult (or earlier — e.g. a pre-visit email)
- Documented in the clinical note — a single line is enough, e.g. “Verbal consent obtained for use of AI scribe (Heidi).”
- Patient has the option to decline without it affecting care
Sample consent script: “I use an AI tool to help with my notes — it listens during our consult and creates a draft that I review and edit before saving. Are you OK with that?”
Obligation 2: Clinician responsibility — AI-assisted, not AI-authored
You read, edit, and sign every note. The AHPRA AI guidelines are clear that AI is a tool that assists the clinician — it does not replace the clinician's professional judgment.
What this looks like in practice:
- The AI-drafted note is reviewed in full before saving
- Any templated phrasing is edited so the note reflects this specific consult
- Clinical reasoning is expressed in your own voice — not the AI's default phrasing
- The note is signed by you (not the AI tool)
If an AHPRA notification is ever lodged against a note, “the AI wrote it” is not a defence. You signed it; you own it.
Obligation 3: Data location — Australian Privacy Principle 8 applies
This is where most clinicians using consumer AI tools (ChatGPT, Claude.ai, Gemini) are unknowingly in breach. Australian Privacy Principle 8 governs offshore disclosure of personal information. ChatGPT (free or Plus), Claude.ai, and Gemini all process inputs on US servers. Pasting any patient health information into those tools is an offshore disclosure.
Under APP 8, offshore disclosure is permitted only when:
- The patient has given explicit, informed consent to the offshore disclosure, OR
- A narrow set of exceptions applies (rare in routine clinical work)
The practical implication: healthcare-purpose-built AU-resident tools (Heidi, Lyrebird) avoid this entire problem. Their data stays on Australian infrastructure; APP 8 doesn't bite.
If you want to use a Tier C tool (ChatGPT, Claude, Gemini) for clinical work, you need either:
- Strip all personal information from the prompt (de-identified data is fine), OR
- Obtain explicit patient consent for the offshore disclosure (rarely practical at scale)
Obligation 4: Documentation integrity — no templating across patients
This is the obligation getting the most regulatory attention in 2026. The NDIA has become noticeably more alert to AI-generated allied health reports as part of fraud-and-assurance work. Templated AI output with generic clinical reasoning and recycled phrasing carries audit risk even when the underlying clinical view is sound.
The integrity test: does your note reflect this specific clinical encounter?
Red flags an auditor (AHPRA, NDIA, Medicare) looks for:
- Same paragraph structure across multiple unrelated patients
- Generic clinical reasoning that could apply to anyone
- Standardised goal language that doesn't reference the specific patient
- Long-form notes that are clearly AI-generated and not edited
The fix is straightforward but requires discipline: edit every note before saving. The AI gives you a draft; you make it specific.
The full framework in one course
Our short course AI in Clinical Practice walks through each of the four AHPRA obligations with worked examples for physiotherapy, osteopathy, GP and naturopathy. Plus the Tier A / B / C tool framework (Heidi vs Lyrebird vs ChatGPT vs Claude), NDIS audit-safe documentation, and the Australian Privacy Principles for clinicians. 3 CPD hours, fully online, certificate on completion. Launching 17 June 2026.
Get the free AI Safety Checklist first