Does my GP have to tell me they're using an AI scribe in our consultation?

Yes. The informed consent framework that applies to clinical care in Australia requires that patients understand what is happening in their consultation, including the use of AI tools. This means your GP should tell you if they are using an AI scribe, explain broadly what it does, and give you the option to decline. In practice, many practices now include this in their intake paperwork or display a sign in the consultation room. If you are unsure whether AI is being used to record or document your consultation, it is entirely appropriate to ask your GP directly.

Is the information recorded by an AI scribe kept private and secure?

This varies by product and practice. AI scribes typically process audio or transcripts of your consultation through the vendor's software — which may involve data being sent to external servers for processing. Different products have different data sovereignty arrangements. Under Australian Privacy Act obligations, your GP's practice is responsible for ensuring that tools they use meet appropriate privacy standards, including storage, access, and deletion of your data. If you have concerns, ask your GP's practice about which product they use, where your data is processed and stored, and what their data retention policy is.

AI scribes in your GP's room: high adoption, thin regulation

What just happened

Government briefing documents obtained under freedom of information laws reveal that AI medical scribes — software that listens to and documents clinical consultations in real time — currently operate with minimal formal regulatory oversight in Australia.

The documents note that “many AI products, such as AI medical scribes, fall outside this framework and have little oversight.” The Therapeutic Goods Administration currently excludes scribes from medical device regulation if they function solely as clinical documentation tools. The TGA’s rationale is that clinical documentation is not itself a medical device function.

The problem identified in a separate TGA review: AI scribes do not always stay within that narrow lane. The TGA found that these tools frequently propose diagnoses or treatments beyond what the clinician identified — crossing from documentation into clinical decision support, which does trigger the medical device framework. Developers have also provided limited transparency about the datasets their models were trained on.

Meanwhile, adoption has accelerated rapidly. A newsGP reader survey from November 2025 found that 40% of responding GPs used AI scribes regularly. Products such as Heidi Health and Lyrebird are among the most widely used. Some vendors are marketing them with claims about potential “30% revenue increases” from efficiency gains.

The Department of Health has confirmed the TGA’s review of widely-used AI medical products is ongoing, with widely-used scribes prioritised.

The both-and

AI scribes solve a real problem in general practice

Clinical documentation is one of the most significant administrative burdens on GPs in Australia. A standard 15-minute consultation can generate 5–10 minutes of note-writing, coding, and referral letter composition — time that either extends well beyond the booked appointment slot or compresses the clinical encounter itself.

AI scribes that accurately transcribe and structure consultation notes address that burden in a way that is practically meaningful. GPs who use them consistently report improved documentation quality and reduced after-hours administrative work. The appeal is not hard to understand.

The ethical logic of the TGA’s current exclusion is also defensible in principle: the GP remains responsible for reviewing, correcting, and signing off on the note before it enters the medical record. The scribe is not making clinical decisions — the GP is. The tool is in the same category as a transcription service or a medical shorthand system.

The regulatory gap appears when the tool does more than that.

When a scribe proposes a diagnosis it hasn’t been asked to make

The TGA’s finding that AI scribes frequently propose diagnoses or treatments beyond what the GP identified is the detail that changes the analysis.

A scribe that proposes a differential diagnosis or treatment option is no longer a documentation tool — it is a clinical decision support system. That function is squarely within the medical device regulatory framework. It requires evidence of safety and efficacy, transparency about training data, and performance validation in the population where it is used. The fact that the proposal appears in a note rather than on a screen labelled “AI diagnosis” does not change its functional character.

The concern is not that AI will replace GP judgement — most GPs are actively reviewing and correcting scribe output. The concern is that an AI-generated clinical suggestion, embedded in a note, can exert subtle anchoring influence on clinical reasoning — a well-documented cognitive phenomenon — without the GP or patient recognising that influence. When the doctor reviews a note and sees “possible iron deficiency anaemia — consider ferritin” typed alongside their dictated history, that suggestion shapes what happens next, even when the GP is confident in their own assessment.

The dataset opacity is a related issue. Clinicians using these tools generally do not know what clinical corpus the model was trained on, whether it over-represents certain demographics, or whether its documentation patterns have been validated in Australian clinical settings. Limited transparency from developers is not a neutral position — it is an information asymmetry that sits on the patient-facing side of the risk.

The RACGP’s Digital Health Chair, Dr Sean Stevens, has been clear that clinical oversight is paramount: “At the end of the day, what’s most important is that the GP or other doctor retains clinical oversight.” The Department of Health has confirmed that informed consent requires patients to understand both the benefits and the limitations of AI tools used in their care.

This is not a theoretical obligation. Patients have a right to know when their consultation is being processed by a third-party AI system, to understand where their data goes, and to decline — without affecting the care they receive. For patients who are disclosing sensitive information (mental health, sexual health, domestic circumstances, substance use), the question of who and what has access to that disclosure is not a minor procedural detail.

2 cents

Next time you see your GP, it is worth checking whether an AI scribe is being used. The question is straightforward: “Is anything recording or transcribing our conversation today?” A good practice will tell you proactively. If they haven’t, asking is entirely reasonable — and the answer should include what happens to your data after the consultation.

The TGA review is the structural piece to watch. If widely-used AI scribes are found to be functioning as clinical decision support tools — which the evidence suggests is already happening — the regulatory framework will need to catch up with the adoption curve. The timeline for that review is not yet published.

For GPs, the specific question is whether their scribe’s output is actively reviewed or implicitly trusted. The note it generates is your responsibility the moment you sign it. What the tool proposed before you reviewed it is where the patient safety question lives.

Verdict: yes — a significant regulatory gap in a rapidly adopted clinical tool, with informed consent implications for every patient whose GP uses one.

Sources cited

newsGP (RACGP) — ‘Little oversight’: officials air AI scribe worries. https://www1.racgp.org.au/newsgp/professional/little-oversight-officials-air-ai-scribe-worries
TGA — Regulation of artificial intelligence as medical devices. https://www.tga.gov.au/resources/resource/guidance/artificial-intelligence-ai-enabled-medical-devices