The next generation of Aotearoa’s GPs will test artificial intelligence-powered virtual consultations as part of a new pilot exploring whether AI can help doctors practise clinical decision-making before facing similar situations with real patients.
The Royal New Zealand College of General Practitioners is partnering with New Zealand-developed platform Gestalt to trial AI-supported simulated consultations with a targeted group of GP registrars, including International Medical Graduates.
Rather than using artificial intelligence to replace doctors or make decisions about real patients, the technology will operate as a training tool, allowing registrars to repeatedly practise consultations, work through possible diagnoses and receive individualised feedback between sessions with medical educators.
For a health system trying to train more GPs while maintaining clinical quality and cultural safety, the pilot could provide an early indication of where AI genuinely helps medical education — and where human judgement must remain firmly in control.
A virtual patient who can be seen again and again
Learning to become a specialist GP requires considerably more than memorising diseases and medications.
General practice often begins with uncertainty.
A patient may arrive with fatigue, pain, dizziness, anxiety or several seemingly unrelated symptoms. The GP must ask the right questions, recognise warning signs, consider multiple possible diagnoses and decide what should happen next.
The College describes specialist GPs as expert generalists who routinely integrate physical health, mental health, family circumstances and social realities while managing clinical uncertainty.
Gestalt is designed to let registrars rehearse that process.
Learners undertake simulated patient consultations before working through differential diagnoses — the range of conditions that could explain a patient’s symptoms — and receiving structured feedback.
That creates something difficult to achieve continuously in conventional medical education: safe, repeatable practice without putting an actual patient at risk.
Practice, feedback, try again
College Chief Executive Toby Beaglehole says the trial is intended to establish whether artificial intelligence adds genuine value within the real GP training environment.
The College’s approach is deliberately cautious.
AI is not being positioned as a replacement for medical educators, clinical judgement or the relationship between doctor and patient.
Instead, registrars could use the technology to gain additional practice between facilitated teaching sessions.
The pilot will examine how frequently registrars engage with it, what their experience is like and what medical educators think about the quality and usefulness of the additional training.
Only after that evaluation will the College consider whether AI-supported learning should play a wider role in its programmes.
New Zealand technology training New Zealand doctors
The technology itself has been developed in Aotearoa.
Gestalt co-founder and chief executive Dr Alastair Dunne, who is also a Fellow of the College, says the platform is intended to supplement established medical education rather than replace it.
Its focus is on giving learners more opportunities to practise the cycle at the centre of clinical development — consultation, reasoning, feedback and reflection.
That distinction is important as artificial intelligence rapidly enters healthcare.
A training simulator represents a considerably different use of AI from allowing an algorithm to independently diagnose or treat a real patient.
If the simulator gets something wrong, a registrar and educator can identify the problem and learn from it.
That makes education one of the environments where the benefits and limitations of medical AI can potentially be tested with considerably lower risk.
More doctors entering GP training
The pilot also comes as Aotearoa attempts to strengthen its primary healthcare workforce.
In February, 225 doctors began specialist training in general practice and rural hospital medicine, including 210 entering the General Practice Education Programme and 15 entering rural hospital medicine training.
The College has repeatedly warned that increasing access to primary healthcare cannot be separated from the number of GPs available to provide it.
Following Budget 2026, it argued that promises to expand patient enrolment and access need to be matched by sufficient workforce capacity.
From February 2027, Health New Zealand will also become a direct-employment option for first-year General Practice Education Programme registrars, another change intended to make the pathway into GP specialisation smoother.
AI training tools will not solve the workforce shortage.
But if they can increase the amount of high-quality practice available to doctors already entering training, they could become another part of strengthening the pipeline.
International doctors are part of the test
Including International Medical Graduates in the pilot could also provide useful information.
Doctors arriving from overseas may have extensive clinical experience while still needing to become familiar with the particular environment of general practice in Aotearoa.
That includes the structure of New Zealand’s health system, referral pathways, local clinical expectations and communication within culturally diverse communities.
Simulated consultations could potentially allow repeated practice of situations specific to the New Zealand setting before or alongside real-world clinical work.
But that also raises an important question for the trial:
How well can an artificial intelligence system understand the cultural and social realities of patients in Aotearoa?
The Māori patient cannot become a stereotype
For hauora Māori, this is where the technology will require particularly careful scrutiny.
A simulated Māori patient cannot simply be an algorithm with a Māori name attached.
Good general practice requires understanding the person, their whānau, their circumstances and the wider factors affecting health.
The College itself says cultural safety, equitable care and improving hauora Māori are essential components of high-quality general practice and patient safety.
That means any AI system used to train future doctors needs to be examined for bias, assumptions and the way it represents different communities.
If the data or scenarios underpinning a system contain stereotypes or gaps, AI can reproduce them at scale.
Conversely, carefully designed simulations could potentially give registrars more opportunities to practise culturally safe communication and reflect on how they engage with different patients and whānau.
The difference will depend heavily on how the technology is designed, evaluated and governed.
Who decides whether the AI’s feedback is right?
There is another fundamental issue.
An AI system providing feedback to a medical learner carries an implied authority.
Registrars need confidence that the advice they receive is clinically sound.
Gestalt says its structured feedback is grounded in clinical sources, while the College’s pilot will evaluate how the technology performs alongside existing education programmes.
That makes educator oversight critical.
Artificial intelligence can produce responses that appear convincing even when they are incomplete or incorrect.
Medical training cannot afford to confuse confidence with accuracy.
The pilot therefore needs to establish not simply whether registrars enjoy using the technology, but whether the feedback genuinely improves clinical reasoning.
Privacy and data sovereignty remain part of the AI conversation
The College has previously identified privacy, data sovereignty, system integration, accuracy and patient confidence among the issues health services need to consider as AI becomes more common.
It has also recognised that artificial intelligence is likely to increasingly support areas such as diagnosis and health data management.
Using simulated rather than real patients reduces some immediate privacy concerns.
But the broader questions remain.
What information is collected about registrars?
How is their performance data stored?
How are simulations developed?
What datasets influence AI responses?
And if Māori knowledge or Māori health scenarios are incorporated, how are Māori data sovereignty and mātauranga Māori protected?
Those questions will become increasingly important if AI-supported education moves from a limited pilot into mainstream medical training.
AI as the practice partner — not the doctor
The significance of this pilot may ultimately be less about the technology itself and more about where the College is drawing the boundary.
The educator remains.
Clinical judgement remains.
The relationship between doctor and patient remains.
AI provides another place to practise.
That distinction could become increasingly important as artificial intelligence moves deeper into medicine.
There is a substantial difference between technology that helps a doctor learn and technology that attempts to replace the doctor’s judgement.
For now, the Royal New Zealand College of General Practitioners is testing the first proposition.
Registrars will practise.
The AI will respond.
Educators will evaluate what happened.
And only then will the College decide whether the technology deserves a bigger place in training Aotearoa’s future GPs.
Because in medicine, the real test of artificial intelligence is not whether it can produce an impressive answer — it is whether using it ultimately helps produce a safer, better and more culturally responsive doctor.
#ArtificialIntelligence #AI #HealthcareAI #MedicalAI #GeneralPractice #GPs #GPTraining #GPRegistrars #RNZCGP #Gestalt #TobyBeaglehole #AlastairDunne #PrimaryCare #HealthWorkforce #MedicalEducation #VirtualPatients #ClinicalTraining #ClinicalReasoning #PatientSafety #HauoraMāori #MāoriHealth #CulturalSafety #MāoriDataSovereignty #MātaurangaMāori #HealthTechnology #DigitalHealth #FutureOfHealthcare #Aotearoa #WaateaNews #RadioWaatea
