This page is designed for people and search engines. Switch to the version for AI systems (LLM) →

· AI Policy Corner · 4 Min. · By Prof. Dr. Oliver Gröne

A watershed moment for AI agents in health care?

Conversational AI agents in healthcare have long been debated — not just as futuristic tools, but as practical helpers. Even in the late 1990s, research indicated that physicians saw value in AI systems for routine inquiries, reminders, and administrative tasks, while expressing skepticism about their ability to address complex clinical needs or provide empathetic care [1,2].

Fast forward nearly three decades and the technical landscape has transformed. On 7 January 2026, OpenAI announced ChatGPT Health, a dedicated health-focused experience. This service allows users to connect personal medical records and wellness app data to enable more personalised, context-aware responses to health and wellbeing questions. While rolling out via a waitlist and unavailable in regions like the EU and UK, it reflects a moment when conversational AI moves beyond generic Q&A toward patient-specific insights [3].

This raises an urgent question: is this a true watershed for AI agents in healthcare — or simply the latest iteration in a long evolution of hype and scrutiny?

From curiosity to everyday use: how patients are already using AI

Patients are not waiting for formal clinical integration to experiment with conversational AI for health concerns. According to OpenAI, over 230 million people globally ask ChatGPT health- and wellness-related questions each week, and millions rely on chatbots daily for health guidance [3]. Real-world usage examples — from both reporting and early deployments of similar tools — include:

  • Daily symptom questions and interpretation: Reports indicate around 40 million Americans use AI chatbots for health advice each day, ranging from interpreting lab results to preparatory questions before appointments.
  • Mental health queries: younger patients increasingly consult generative AI for emotional support and psychiatric symptoms — raising ethical and clinical concerns about self-directed therapy without oversight.
  • Health record interpretation: Early adopters of ChatGPT Health are connecting their records and wellness data to receive tailored explanations about medication regimens, lab trends, or what to ask their clinician — users appear to value to quick feedback over privacy concerns.

This rapid uptake reflects a broader pattern identified in the academic literature: patient acceptance and trust in AI tools — including for sensitive health topics — is growing, even as understanding of risks remains incomplete [4].

Shifting focus: from “can AI know?” to “how should it be used?”

Much of the research and discourse has moved past whether conversational AI can recall medical facts accurately. Instead, the critical questions now are how to ensure safe, appropriate application in clinical contexts and how to preserve humane, patient-centred care [5-9].

Academic evidence shows that conversational AI can be effective when supervised and integrated thoughtfully. For example, a controlled study of a physician-supervised AI agent in a real-world service found that patients reported higher clarity and satisfaction with AI-assisted conversations than with standard care, with clinician oversight ensuring safety [10]. At the same time, reviews highlight limitations: empathy remains a complex human attribute – but even in terms of empathy AI can already outperform humans (based on text-only interactions) [2].

Meanwhile, rigorous research into clinician acceptance shows mixed attitudes: performance expectancy and facilitating conditions support AI adoption, but clinician hesitancy, legal concerns, and contextual factors — including workflow and regulation — significantly influence whether and how these tools are used [11].

Clinicians and systems: keeping pace with rapid AI adoption

The pace at which patients are adopting conversational AI poses both opportunities and challenges for clinicians and health systems.

  • Professional bodies and health systems must invest in formal education on AI use in clinical contexts. Clinicians will need new competencies in human-AI collaboration, including understanding AI limitations, interpretability, and safe escalation pathways. Without this, clinicians risk being outpaced by tools their patients already use.
  • Rather than viewing AI as external, health systems are already embedding its potential within workflows. For example, separate but related tools (e.g., ChatGPT for Healthcare) help clinicians draft documentation and summaries, potentially reducing administrative burden and allowing more focus on direct patient care [12].
  • Given evidence that users may trust AI advice even when incorrect, health systems will need real-time monitoring of outcomes and safety signals, supported by clinician review and reporting frameworks. This is vital to guard against misinformation and ensure AI augments — not undermines — clinical judgement.
  • Health systems must develop clear governance around AI usage — including data privacy, security, accountability, and escalation protocols. ChatGPT Health’s emphasis on encryption and user-controlled data highlights privacy concerns at the center of trust debates, but policy frameworks must be equally robust and internationally coherent.

Reasons to be pessimistic?

Public healthcare systems have historically struggled to keep pace with digital health innovation, and conversational AI agents risk amplifying this long-standing gap. Past experience with electronic health records and telemedicine shows that adoption delays were rarely due to technical limitations, but rather to institutional inertia, fragmented governance, workforce readiness gaps, and regulatory uncertainty—challenges that have been particularly acute in publicly funded systems.

In contrast, private-sector technology firms can deploy and iterate AI tools at speed, shaping patient behaviour long before health systems formally respond. As indicated above, evidence already shows patients using conversational AI for symptom interpretation, mental health support, and understanding test results, while clinicians often lack formal training in human–AI collaboration, interpretability, and escalation pathways. Without rapid investment in clinical competencies, workflow integration, real-time monitoring, and adaptive governance, public institutions risk being outpaced — left reacting to AI-mediated patient decisions rather than proactively shaping safe, equitable, and trustworthy use of AI in care [13].

A watershed — only with the right supports

So, has conversational AI in healthcare reached a watershed? The short answer is: potentially — but only if deployment is accompanied by robust governance, clinician preparedness, and evidence-based integration.

ChatGPT Health and similar systems mark a transition from experimental chatbots to tools tightly coupled to personal health data and daily patient decision-making. But technology alone won’t define the impact — practice, policy, and professional readiness will.

This moment could indeed be transformative — not because AI can answer questions better, but because patients are already using it, systems must respond, and clinicians must lead the integration in ways that reinforce trust, safety, and equitable care.

References

  1. Palanica A, Flaschner P, Thommandram A, et al. Physicians’ Perceptions of Chatbots in Health Care: Cross-Sectional Web-Based Survey. J Med Internet Res 2019;21(4):e12887 doi: 10.2196/12887
  2. Howcroft A, Bennett-Weston A, Khan A et al. AI chatbots versus human healthcare professionals: a systematic review and meta-analysis of empathy in patient care, British Medical Bulletin, Volume 156, Issue 1, December 2025, ldaf017, https://doi.org/10.1093/bmb/ldaf017
  3. https://openai.com/index/introducing-chatgpt-health/ [accessed 25.1.2026]
  4. Kauttonen J, Rousi R, Alamäki A. Trust and Acceptance Challenges in the Adoption of AI Applications in Health Care: Quantitative Survey Analysis. J Med Internet Res 2025;27:e65567, doi: 10.2196/65567
  5. Sunshine A, Honce GH, Callen A et al. Evaluating the Quality and Understandability of Radiology Report Summaries Generated by ChatGPT: Survey Study, JMIR Form Res 2025;9:e76097, DOI: 10.2196/76097
  6. Liao W, Li M, Ma C et al. Developing a Quality Evaluation Index System for Health Conversational Artificial Intelligence: Mixed Methods Study. J Med Internet Res 2026 (Jan 19); 28:e83188
  7. Meyer A, Karay Y, Steinbicker A et al. Performance of DeepSeek-R1, ChatGPT (GPT-o3-mini), and Gemini 2.0 Flash on German Medical Multiple-Choice Questions: Comparative Evaluation. JMIR Form Res 2025 (Dec 18); 9:e77357
  8. Pornwattanakavee S, Leelakanok N, Todsarot T et al. Effectiveness of ChatGPT, Google Gemini, and Microsoft Copilot in Answering Thai Drug Information Queries: Cross-Sectional Stud. JMIR AI 2025;4:e79751, doi: 10.2196/79751
  9. Aliyeva A, Alaskarov E. AI at the Helm: Evaluating Claude 3.5 Sonet and ChatGPT-4.0 in Tympanoplasty Management. Otol Neurotol. 2026 Jan 5. doi: 10.1097/MAO.0000000000004812.
  10. Lizée A, Beaucoté PA, Whitbeck J et al. Conversational Medical AI: Ready for Practice. arXiv:2411.12808 [cs.AI]
  11. Scipion CEA, Manchester MA, Federman A, et al. Barriers to and facilitators of clinician acceptance and use of artificial intelligence in healthcare settings: a scoping review. BMJ Open 2025;15:e092624. doi:10.1136/bmjopen-2024-092624
  12. ChatGPT for Healthcare – Resource | OpenAI Academy. https://academy.openai.com/public/clubs/work-users-ynjqu/resources/chatgpt-for-healthcare?utm_source [accessed 25.1.2026]
  13. Angus DC, Khera R, Lieu T, et al. AI, Health, and Health Care Today and Tomorrow: The JAMA Summit Report on Artificial Intelligence. JAMA. 2025;334(18):1650–1664. doi:10.1001/jama.2025.18490

Related Articles

Related articles