Ahpra AI guidance: what it means for paediatric practice
Ahpra has no new AI rules. Here is how its AI guidance applies to OTs, psychologists, paediatricians and speech pathologists, with a consent script.
8 min read
Ahpra and the National Boards have not created new rules for AI. Their guidance, "Meeting your professional obligations when using AI in healthcare", explains how the existing codes of conduct apply when you use it. The core expectations are these: you stay accountable and apply human judgement to any AI output; you understand the tool you are using; you are transparent with patients; you get informed consent when AI uses their personal data, which is generally needed for AI scribes; and you meet your privacy, Therapeutic Goods Administration (TGA) and bias obligations. This guide covers what that means in a paediatric practice. It is general information, not legal advice.
Who the guidance applies to in a paediatric practice
Ahpra (often written AHPRA) works with the National Boards to regulate registered health practitioners under the National Registration and Accreditation Scheme. In a paediatric practice, its guidance reaches some of your team and not others.
- Occupational therapists are Ahpra-registered. They are covered by the Shared Code of conduct used by 12 National Boards, including occupational therapy. Section 8.3 of that code covers health records.
- Psychologists are Ahpra-registered. The shared code does not apply to psychology. The Psychology Board has its own code of conduct, which took effect on 1 December 2025. The Australian Psychological Society (APS) has also announced professional practice guidelines on AI and emerging technologies.
- Paediatricians are medical practitioners and are Ahpra-registered. The Medical Board's Good medical practice applies to them.
- Speech pathologists are not Ahpra-registered. The profession is self-regulated through Speech Pathology Australia (SPA), and "speech pathologist" is not a protected title. SPA's position statement "Artificial intelligence in speech pathology: Ethical considerations" (May 2024) sets out similar expectations on consent, confidentiality, accountability and critique of AI output.
If you run a mixed practice, set one AI policy that meets the strictest of these. One policy is easier to teach and audit than four.
What does Ahpra's AI guidance say?
The guidance covers five areas.
Accountability. The guidance says practitioners "must apply human judgment to any output of AI". TGA approval of a tool does not remove that duty.
Understanding the tool. Review the product information: how it was trained and tested, what it is intended for and its limitations. Find out how your data is used to retrain the AI and where it is stored.
Transparency. Tell patients when you use AI. How much detail you give depends on the use.
Informed consent. Involve patients when AI uses their personal data, and ideally note their response in the health record. For generative AI scribes, informed consent will generally be needed.
Ethical and legal obligations. Meet privacy and health records law, and watch for bias, including for Aboriginal and Torres Strait Islander people and people from diverse backgrounds. Meet TGA and state requirements, follow your employer's AI governance, and check that your professional indemnity insurance covers AI use.
Ahpra also says public generative AI tools carry their own risk: personal information you enter may become public, so check whether patient data is used to train the model. The Office of the Australian Information Commissioner (OAIC) goes further and recommends, as best practice, that organisations "do not enter personal information, and particularly sensitive information, into publicly available generative AI tools".
Do I need consent to use an AI scribe?
Yes, generally. Ahpra says informed consent will generally be needed for generative AI scribes, and it suggests noting the patient's response in the health record. It also warns there "may be criminal implications if consent is not obtained before recording". That refers to state and territory surveillance or listening device laws, which differ between jurisdictions, so check the law where you practise.
In paediatrics, the person who consents is usually the parent or guardian with legal authority. The child should still hear, in age-appropriate terms, what is happening and why. How consent works for older children and mature minors varies, so check the rules that apply to your discipline and location before you rely on a parent's consent alone.
Psychologists should note the APS position as summarised in an APS article: written informed consent before AI touches client data, consent specific to each tool, and the line that "Blanket consent for 'AI use' is not sufficient". SPA adds that a client's decision to decline AI should not reduce their access to services of equal quality. Offer a scribe-free session as a real option.
Example consent script and record entry
This is an example to adapt for your own practice, not legal wording. The child is fictional.
"Before we start, I'd like to use a tool that listens to our session and drafts my notes. I check and correct everything it writes before I sign it, and the notes stay part of Mia's record like any others. [Explain what happens to the audio and the transcript, for example whether the audio is kept and for how long.] You can say no, and Mia will get exactly the same session and the same quality of care. You can also change your mind at any point, and I'll switch it off. Do you consent to me using it today? I'll also tell Mia, in words that make sense to her, that I'm using a helper to write notes so I can pay more attention to her."
An example record entry: "Parent consented verbally to ambient scribe use for this session; child informed in age-appropriate terms."
Who is responsible if the AI gets it wrong?
You are. Ahpra's guidance is direct that the practitioner is responsible for checking the accuracy and relevance of records created with generative AI.
Ahpra's second case study shows why. A scribe adds a plausible diagnosis, one the practitioner never made, to the notes and to a referral letter. The lessons are hallucination and omission, the risk that a tool suggesting diagnoses may be a medical device, and the point that the practitioner remains responsible. In a paediatric setting a stray diagnosis in a referral letter can follow a child through school, funding and later care, so reading every line before you sign is part of the clinical task.
SPA asks the same of speech pathologists: AI-generated information needs "detailed critique" for currency, accuracy and relevance, and AI should not replace professional decision making.
Is my AI tool regulated by the TGA?
It depends on what the tool does. TGA guidance on digital scribes (August 2025) says scribes that only record, transcribe and summarise, without adding clinical content or suggestions, are generally not medical devices. A scribe with diagnostic or treatment-recommendation features may be a medical device that needs inclusion on the Australian Register of Therapeutic Goods (ARTG) or an exemption. Disclaimers do not override features. Clinicians can check the ARTG themselves.
Two things follow. First, ask vendors in writing which features their tool has and whether any is a TGA-regulated device. Second, remember that even a TGA-approved tool does not relieve you of your duty to apply judgement to its output.
A practical checklist for your practice
- Write an AI policy. Name the approved tools, who may use them, and what may never be entered into a public generative AI tool.
- Ask vendors the due diligence questions. How was the tool trained and tested? What is it intended for and what are its limits? Is client data used to retrain it? Where is data stored, including offshore? Does it have diagnostic or treatment features that could make it a medical device?
- Use a consent script and record the response. Adapt the example above, and note the answer in the health record each time.
- Set a review-before-sign rule. Nothing AI-drafted enters the record or leaves the practice until the clinician has read, corrected and signed it.
- Mark AI-assisted content where your professional body requires it. The APS guidance summary says to clearly identify AI-produced content, and SPA asks for transparent representation of AI use.
- Check your indemnity insurance. Ask your insurer in writing whether AI use is covered.
- Update your privacy policy before 10 December 2026. From that date, automated decision-making transparency (APP 1.7 to 1.9) applies where a computer program makes, or substantially assists, a decision that could significantly affect someone's rights or interests. The OAIC says generative AI counts as a computer program and that a decision can be in scope even when a human reviews the output. Ask vendors for clear high-level information to help you write the policy.
- Review the policy when tools update. The OAIC says due diligence is not "set and forget", and Ahpra says its own guidance will be updated as AI changes.
In short
- Ahpra has no new AI rules; its guidance applies the existing codes of conduct.
- OTs, psychologists and paediatricians are Ahpra-registered. Speech pathologists follow SPA's position statement. Mixed practices should meet the strictest standard.
- Informed consent is generally needed for AI scribes. In paediatrics, that means the parent or guardian with legal authority, plus an age-appropriate explanation to the child.
- You remain responsible for every word you sign, and a scribe with diagnostic features may be a TGA-regulated medical device.
- Update your policies, your privacy policy and your insurance check as your tools change.
FAQ
Has Ahpra released new rules on AI?
No. Ahpra and the National Boards say their guidance explains how existing codes of conduct apply to AI. Ahpra has said it will review and update the guidance as AI changes, so check the page for the current version.
Do speech pathologists need to follow Ahpra's AI guidance?
Speech pathologists are not Ahpra-registered, so the guidance does not apply to them directly. Speech Pathology Australia's May 2024 position statement on AI sets out similar expectations, including informed consent, confidentiality, accountability and critique of AI output.
Do I need consent to use an AI scribe with a child?
Ahpra says informed consent will generally be needed for generative AI scribes. In paediatrics, that usually means consent from the parent or guardian with legal authority, plus an explanation to the child suited to their age. Rules for older children and mature minors vary, so check the requirements where you practise.
Is an AI scribe a medical device?
Not usually. TGA guidance says scribes that only record, transcribe and summarise, without adding clinical content or suggestions, are generally not medical devices. A scribe with diagnostic or treatment-recommendation features may be, and would need ARTG inclusion or an exemption.
Syntropic Health's ambient scribe and clinical notes draft inside the case, and the clinician reviews, edits and signs everything before it becomes part of the record.
Sources
Get new resources by email
New research and practical guides for paediatric practice.
Share this article
More resources
Take the next step
Book a demo, or explore the platform, trust page and research lab.