What does “a better health system” actually mean: fewer queues and fuller wards… or a calmer, more coherent experience of care from first symptom to follow-up?
Australia already tries to measure what matters via the Australian Health Performance Framework (AHPF), which deliberately expands beyond narrow throughput metrics to include safety, continuity of care, accessibility, appropriateness, and (crucially) equity—linking “determinants of health” to “health system performance” to “health status” (Australian Institute of Health and Welfare [AIHW], 2024a). The framework diagram many of us have seen is explicitly attributed as “adapted from” the National Health Information and Performance Principal Committee’s 2017 work (AIHW, 2017). That matters, because it implies something subtle: the framework is not a static scoreboard; it is a navigation tool that governments can (and arguably must) keep evolving as new forms of data become available (AIHW, 2024b).
And this is where the tension sharpens. Most Western countries are ageing, and end-of-life care concentrates costs in a way that can feel ethically unavoidable yet operationally destabilising. AIHW estimates that, across a major study period, around 8% of spending on four health service types was for people in their last year of life (AIHW, 2022a). Even more pointedly—and personally resonant for my family—AIHW reports that the average annual cost in the last year of life was highest for people who died from bowel (colorectal) cancer ($40,700 per person) (AIHW, 2022b). Years ago, I remember asking my brother Chip (a colorectal surgeon) a naïve futurist’s question: “Why don’t we just do a Star Trek–style whole-body scan and catch everything early?” His answer was clinical realism: because we all have “defects”; most are incidental; and knowing them often increases anxiety without improving diagnosis or outcomes. In other words, more data does not automatically create more health—unless it is organised, contextualised, and used with judgement.
So what changes when AI becomes unusually good at organising the messy middle of healthcare: the unstructured, the fragmented, the narrative?
In January 2026, OpenAI launched “ChatGPT Health”, explicitly positioning it as a dedicated space for medical and wellness questions that can work with a user’s health data—but “not designed to replace your doctor” (OpenAI, 2026; Silberling, 2026). The headline statistic is startling: Reuters reported that more than 230 million health-related questions are asked on ChatGPT every week (Reuters, 2026). If you’re wondering why this matters now, that is the answer: demand is already here, at global scale, before any formal integration with health systems. ChatGPT Health’s promise is essentially narrative coherence: users can upload records and connect wellness apps (for example, Apple Health and other trackers), then ask questions that turn scattered results, notes, and trends into something closer to an intelligible story—preparing for appointments, interpreting results, and tracking patterns over time (OpenAI, 2026; Reuters, 2026). OpenAI also stresses compartmentalisation and privacy controls—Health is “separate”, and Health chats/files/memories are not used to train foundation models (OpenAI Help Centre, 2026).
This is not just a “new feature”; it is a signpost for a broader transformation. AI’s practical superpower in health is not mystical diagnosis—it is structure: converting unstructured notes into timelines, extracting key entities (medications, comorbidities, allergies), reconciling conflicting lists, highlighting gaps, and presenting a patient-friendly brief. It also acts as a cognitive prosthetic. Humans—clinicians and patients alike—are vulnerable to well-documented biases: availability bias (recent dramatic cases loom larger), anchoring (first impressions stick), confirmation bias (we over-weight evidence that fits), and omission bias (we prefer “doing nothing” even when action is warranted). Properly designed, AI can counter some of these by forcing explicitness:
- What evidence supports this?”
- “What would change your mind?”
- “What’s missing?”
Yet it can also amplify biases if it simply mirrors our framing. The point is not that AI is “objective”; it is that it can be engineered to be systematically questioning.
If we already tried shared digital health records, why would “this time” be different?
It begins with trust—and Australia’s history with national identifiers shows how hard that is to earn. The Hawke Government’s proposed Australia Card (1985–87) triggered intense backlash over surveillance and privacy, and ultimately collapsed after political and public resistance (Clarke, 1986/updated; Australian Privacy Foundation, n.d.). By contrast, countries like Estonia built digital governance around a widely adopted state digital identity, with unique identifiers underpinning access to services—including health records—within a higher-trust digital culture (e-Estonia, n.d.; Tartu City Government, 2025). Australia did pursue personally controlled electronic health records (with the PCEHR launching in 2012) and later iterations of My Health Record, supported by legislation and privacy oversight (Pearce & Bainbridge, 2014; Office of the Australian Information Commissioner, n.d.; Australian Digital Health Agency, n.d.).
But the lived experience for many Australians has been patchy: uneven clinician uptake, inconsistent data completeness, and a persistent sense that “my record” is not quite my narrative.
ChatGPT Health hints at a different adoption pathway: consumer-pulled rather than system-pushed. People are already asking the questions. The interface meets them where they are, then works backwards into integration. That creates a new policy dilemma: if citizens build their own health narratives via AI tools before governments or health systems standardise the plumbing, regulation risks becoming a reactive clean-up job rather than proactive design.
What does this mean for mental health—where continuity is often the whole problem?
Between 2011 and 2016, the Young and Well Cooperative Research Centre (Young and Well CRC) explored how digital technologies could improve youth mental health and wellbeing, convening researchers, practitioners, and service organisations (Western Sydney University, n.d.). Contemporary reporting indicates Commonwealth-scale investment consistent with the CRC model (for example, a ~$27m cash investment is widely cited in public commentary) (ReachOut, 2011; Sweet, 2023). My consulting work in that period proposed a broadly similar systems architecture ambition: improving continuity for young people who otherwise “bounced” between specialist organisations (e.g., headspace, ReachOut, Beyond Blue) with limited shared visibility of a person’s journey. In practice, however, there was limited political and organisational appetite for a shared platform. Each organisation operated within distinct funding arrangements, governance constraints, and service logics, which shaped priorities and reduced incentives for deep interoperability. Even where the need for standardised integration was understood, health-related APIs and integration patterns were still emergent, and cross-organisation alignment would have required a level of shared architecture—technical, contractual, and cultural—that the ecosystem was not yet positioned to carry. The vision was right; the integration burden was heavy—too many systems, too many incentives, too little coordinating infrastructure.
Yet one outcome of that broader period was tangible service expansion. headspace’s reporting shows a network of 100 centres by 2016–17 (headspace, 2017), and subsequent annual reporting points to a larger national footprint (headspace, 2025). Meanwhile, Australia has watched public mental health staffing strain in real time; for example, the NSW public psychiatrist resignation dispute underscored fragility in specialist capacity and continuity (Australian Broadcasting Corporation, 2024; Guardian Australia, 2025). A recent Lowitja Institute review further reinforces that Aboriginal and Torres Strait Islander social and emotional wellbeing is a culturally grounded, holistic and relational concept—shaped by connections to family, community, culture and Country—so any AI-enabled “shared responsibility” agenda must be strengths-based and culturally safe rather than narrowly individualised (Dudgeon et al., 2025).
This leads to a provocative proposition: in the future, mental wellbeing cannot be treated as something delivered only by governments and clinicians. It will remain their responsibility—but it will increasingly become our responsibility as individuals and communities, supported by tools that make self-monitoring, earlier help-seeking, and better preparation for care more feasible. The workplace trend is a micro-signal: just as organisations distribute safety roles (first aid officers, fire wardens), many now formalise mental health first aid roles to guide colleagues towards support and reduce stigma (CPSU, n.d.; Mental Health First Aid Australia, n.d.).
What if everyday health became as visible—and as actionable—as today’s weather?
Consider what we already accept as “live feeds”: stock markets update constantly, shaping decisions at scale; weather apps bundle multiple signals—temperature, wind, UV, air quality—into a quick read that changes behaviour.
Imagine a near-future dashboard that adds two more layers:
- Your personal health signals (sleep debt, resting heart rate drift, activity, stress proxies)
- Your community health signals (de-identified trends by postcode, workplace, or campus)
The metaphor here is the fire danger rating: simple categories that compress complex risk into clear action guidance. That is not “dumbing down”; it is designing for action under uncertainty. The old management line—often attributed to Peter Drucker—is that “what gets measured gets managed” (the attribution is debated, but the idea persists for a reason). In public policy terms, “counting what counts” is how societies steer social change. If daily, comprehensible health metrics became normal, prevention could become a lived habit rather than an annual pamphlet—especially when paired with passive sensing via watches, phones, and eventually more continuous biometrics.
Big tech is already positioning for this ecosystem. Apple’s health strategy has steadily moved from wellness tracking towards regulated features (for example, ECG and irregular rhythm notifications in Australia as ARTG-listed devices) (Apple, 2021). Google continues to frame “Google for Health” as a platform ambition, while Fitbit’s AI coaching direction points to personalised behaviour change at scale (Google, n.d.; Fierce Healthcare, 2025). Microsoft is betting on clinician workflow via ambient documentation and AI copilots to reduce administrative burden (Microsoft, n.d.). ChatGPT Health sits between these worlds: consumer narrative on one side, workflow automation and evaluation on the other.
To be credible, this future needs rigorous safety science. OpenAI’s HealthBench is one example of an attempt to evaluate model performance in realistic health scenarios shaped by clinician judgement—an important step towards measurement discipline rather than marketing optimism (OpenAI, 2025).
So, as we enter a New Year of resolutions, here are 3 predictions related to Health ?
- Personal health stewardship will become normalised: people will maintain a simple, portable “health ledger” (baselines, meds, allergies, key events, goals), increasingly auto-updated from devices, and shared deliberately—turning “ownership of health” from a slogan into an everyday practice of continuity.
- Pre-visit medicine will become routine: AI-assisted preparation that helps patients bring cleaner histories, clearer questions, and better follow-through—improving comfort and comprehension even when outcomes are unchanged.
- Preventive maintenance will shift from slogans to systems: passive biometrics plus trend interpretation will normalise earlier intervention—especially for cardiometabolic risk and mental wellbeing—provided privacy and consent stay central.
- Community health will become a shared civic object: de-identified dashboards that help local services, workplaces, and councils respond to emerging patterns—much as communities already respond to heatwaves, smoke, or infectious disease signals.
If that sounds ambitious, good. The alternative is to keep judging “success” by bed blocks and waiting times while patients experience care as fragmented and anxiety-inducing.
Momentum is not merely technological; it is cultural. If we treat health like a shared asset—something we steward personally and collectively—then the New Year’s resolution stops being a private guilt ritual and becomes a social flywheel. And perhaps that is the real catalyst: not AI replacing clinicians, but AI making it easier for all of us to participate in a healthier, more accountable, more humane system.
Ian Farmer
References
Apple. (2021, May 26). ECG app and irregular rhythm notification now available on Apple Watch. https://www.apple.com/au/newsroom/2021/05/ecg-app-and-irregular-rhythm-notification-now-available-on-apple-watch/
Australian Broadcasting Corporation. (2024, December 22). Minns government refuses to back down, increases locum rates amid NSW public psychiatrists crisis. https://www.abc.net.au/news/2024-12-23/private-doctors-crisis-rates-nsw-public-psychiatrists/104758242 (ABC)
Australian Digital Health Agency. (n.d.). My Health Record legislation and governance. https://www.digitalhealth.gov.au/about-us/policies-privacy-and-reporting/my-health-record-legislation-and-governance (digitalhealth.gov.au)
Australian Institute of Health and Welfare. (2017). Australian Health Performance Framework: Conceptual framework (diagram) [PDF]. https://www.aihw.gov.au/getmedia/0ac81e61-0558-4ee0-8575-dba751d6772f/ahpf-framework-diagram.pdf.aspx (AIHW)
Australian Institute of Health and Welfare. (2022a). The last year of life: Patterns in health service use and costs – Key findings. https://www.aihw.gov.au/reports/life-expectancy-deaths/the-last-year-of-life-health-service-use-patterns/contents/key-findings (AIHW)
Australian Institute of Health and Welfare. (2022b). The last year of life: Health service use and costs by cause of death. https://www.aihw.gov.au/reports/life-expectancy-deaths/the-last-year-of-life-health-service-use-patterns/contents/health-service-use-and-costs-by-cause-of-death (AIHW)
Australian Institute of Health and Welfare. (2024a). Australia’s health performance (hub). https://www.aihw.gov.au/reports-data/australias-health-performance (AIHW)
Australian Institute of Health and Welfare. (2024b). Learn more about the framework. https://www.aihw.gov.au/reports-data/australias-health-performance/learn-more-about-the-framework (AIHW)
Australian Privacy Foundation. (n.d.). Campaign against the Australia Card. https://privacy.org.au/about/history/davies0402/ (privacy.org.au)
Clarke, R. (n.d.). Roger Clarke’s ‘Australia Card’ (background and analysis). https://www.rogerclarke.com/DV/OzCard.html (rogerclarke.com)
Community and Public Sector Union. (n.d.). Mental Health First Aid in the workplace. https://www.cpsu.org.au/CPSU/Content/Safety_First_Network/WHS_KYR/mental_health_first_aid_workplace.aspx (CPSU)
Dudgeon, P., Gibson, C., Walker, R., Bray, A., Agung-Igusti, R., Derry, K., Gray, P., McPhee, R., Sutherland, S., & Gee, G. (2025). Social and emotional wellbeing: A review. Lowitja Institute. https://doi.org/10.48455/4f9e-3v29
e-Estonia. (n.d.). ID-card / e-Identity. https://e-estonia.com/solutions/estonian-e-identity/id-card/ (e-Estonia)
Fierce Healthcare. (2025, August 25). Health Tech roundup: Google unveils AI health coach for Fitbit… https://www.fiercehealthcare.com/health-tech/health-tech-roundup-google-unveils-ai-health-coach-fitbit-ambience-debuts-chart-chat (Fierce Healthcare)
Google. (n.d.). Google for Health. https://health.google/ (Google Health)
Guardian Australia. (2025, October 3). Public hospital psychiatrists to receive 20% pay rise after winning dispute with NSW Health. https://www.theguardian.com/australia-news/2025/oct/03/public-hospital-psychiatrists-to-receive-20-pay-rise-after-winning-dispute-with-nsw-health (The Guardian)
headspace. (2017). Annual Report 2016–17 [PDF]. https://headspace.org.au/assets/Uploads/headspace-Annual-Report-for-web2.pdf (headspace.org.au)
headspace. (2025). Annual Report 2025 [PDF]. https://headspace.org.au/assets/Reports/Annual-reports/Headspace-Annual-Report-2025_WEB.pdf (headspace.org.au)
Mental Health First Aid Australia. (n.d.). About MHFAiders. https://www.mhfa.com.au/about-us/mhfaiders (Mental Health First Aid Australia)
Microsoft. (n.d.). Dragon Copilot | Microsoft for Healthcare. https://www.microsoft.com/en-us/health-solutions/clinical-workflow/dragon-copilot (Microsoft)
Office of the Australian Information Commissioner. (n.d.). My Health Record. https://www.oaic.gov.au/privacy/privacy-legislation/related-legislation/my-health-record (OAIC)
OpenAI. (2025, May 12). Introducing HealthBench. https://openai.com/index/healthbench/ (OpenAI)
OpenAI. (2026, January). Introducing ChatGPT Health. https://openai.com/index/introducing-chatgpt-health/ (OpenAI)
OpenAI Help Centre. (2026). What is ChatGPT Health? https://help.openai.com/en/articles/20001036-what-is-chatgpt-health (OpenAI Help Center)
Pearce, C., & Bainbridge, M. (2014). A personally controlled electronic health record for Australia. Journal of the American Medical Informatics Association. https://pmc.ncbi.nlm.nih.gov/articles/PMC4078271/ (PMC)
ReachOut. (2011, March 7). Multi-million dollar Cooperative Research Centre to improve the wellbeing of young Australians. https://about.au.reachout.com/blog/multi-million-dollar-co-operative-research-centre-to-improve-the-wellbeing-of-young-australians (About ReachOut Australia)
Reuters. (2026, January 7). OpenAI launches ChatGPT Health to connect medical records, wellness apps. https://www.reuters.com/business/healthcare-pharmaceuticals/openai-launches-chatgpt-health-connect-medical-records-wellness-apps-2026-01-07/ (Reuters)
Silberling, A. (2026, January 7). OpenAI unveils ChatGPT Health, says 230 million users ask about health each week. TechCrunch. https://techcrunch.com/2026/01/07/openai-unveils-chatgpt-health-says-230-million-users-ask-about-health-each-week/ (TechCrunch)
Sweet, M. (2023, August 30). Responding to concerns about University of Sydney digital mental health platform. Croakey. https://www.croakey.org/responding-to-concerns-about-university-of-sydney-digital-mental-health-platform/ (Croakey Health Media)
Tartu City Government. (2025, December 31). Personal identification code. https://tartu.ee/en/personal-identification-code (Tartu linn)
Western Sydney University. (n.d.). Young and Well Cooperative Research Centre (past project). https://www.westernsydney.edu.au/ics/projects/past-projects/young-and-well-cooperative-research-centre (westernsydney.edu.au)