AI clinical notes in Australia: what allied health should know

Notes are the job that is still waiting for you at the end of a full day. A busy diary leaves a pile of documentation for the evening, and AI tools that turn a spoken summary into a structured note look like an obvious answer. They can genuinely help. They also move some of your professional and legal obligations into new places, and it is worth knowing where before you switch one on.

This post covers the questions an Australian allied health practice should settle first: consent, accuracy, sign-off, where the data goes, and how long the record must be kept. It is general information, not legal advice, and guidance in this area is still developing.

Your professional obligations do not change

Ahpra and the National Boards have published guidance on using AI in health practice. It frames AI through the codes of conduct you already work under, around accountability, understanding, transparency, informed consent, and ethical and legal issues. In plain terms: you remain accountable for the care you provide and the records you keep, you should understand what a tool does and where its limits are before you use it, you should be open with patients about how it is used, you should involve them in decisions about tools that use their information, and privacy and other legal duties still apply. Read the current version on the Ahpra website, and check whether your professional association has more specific guidance for your discipline.

In practice, an AI-drafted note becomes your note the moment you accept it. If it records a test you did not perform, or a finding you ruled out, the error is yours. The tool is a typist and an editor, not a co-author.

Consent and transparency

AI note tools tend to work in one of two ways, and they raise different consent questions.

• Recording the consultation. Tools that listen to and record a session raise the biggest questions. Patients should know a recording is being made, why, where it goes and when it is deleted, and should be able to say no without it affecting their care. Laws on recording conversations differ between states and territories, so check the rules where you practise.
• Drafting from your own words. A tool that takes the summary you dictate or type after the session and arranges it into a note is closer to a dictation service. You should still tell patients, in your privacy policy and at intake, that you use software, including AI, to help prepare their records.

Either way, write down what you tell patients and record their consent on the file. A consent you cannot show is a consent that is hard to rely on.

Accuracy and clinician sign-off

AI drafting tools are good at structure and wording, and they can still get facts wrong. The errors to watch for are quiet ones: a side recorded as left instead of right, a repetition count changed, a symptom the patient denied written up as present, or a plan that sounds more definite than what you said.

1. Read every draft in full before you accept it. Do not skim the parts that look routine.
2. Check measurements, sides, numbers and negatives line by line.
3. Make sure the plan reflects what you actually agreed with the patient.
4. Sign off only when the note says what you would have written yourself.
5. Keep the ability to see what changed. If a note is edited after sign-off, the earlier version should still exist.

Where the data goes, and how long you keep it

Health information is sensitive information under the Privacy Act 1988 and the Australian Privacy Principles, and some states and territories add their own health records laws. That leaves two practical questions for any AI tool.

First, where is the information processed? Many AI services run outside Australia. That is not automatically prohibited, but the Australian Privacy Principles set out what you must do before personal information is disclosed overseas, and your privacy policy should say so honestly. Ask the vendor which parts of the service run where, and get the answer in writing.

Second, what is kept? The finished note is part of the health record and must be kept for the period required where you practise, and some laws set longer periods for children's records. Audio recordings and intermediate drafts are a separate question: find out whether the tool stores them, for how long, and whether they become part of the record. Confirm your obligations with the OAIC and, where one applies, your state or territory health records regulator.

How AspireAi handles notes

AspireAi, the assistant built into AspireOS, drafts and never signs. A clinician dictates or pastes a summary of the consult, and the AI note sorts it under the practice's own note headings for review. It does not record the consultation. Dictation uses the browser's own speech recognition, one field at a time, so it is not available in every browser. Nothing enters the note until the clinician accepts it, notes are completed and locked, and later changes are kept as versions. AspireAi does not diagnose or choose treatments, and AI activity is part of the audit trail.

Practice data is hosted in Sydney, AspireAi runs on OptiTide's custom AI, hosted in Sydney, and AspireOS's privacy policy lists every service involved and what each one receives. That is the kind of list to look for in any vendor's documents. Our security page sets out the rest of how data is handled, the reasoning behind the draft-first design is in why your practice-management system should draft, not decide, and the feature list shows where the AI note sits alongside the rest of the record.

If you would like to see how a dictated summary becomes a draft note, and what the clinician checks before it is saved, start a free trial.