The AI Parallel: Integrating Medical Records with ChatGPT and Anthropic

2026, Medicine, Science Newsletter / By
August 7, 2026

As access to healthcare professionals becomes increasingly limited, many patients are turning to alternative sources of medical information. Large language models developed by companies such as OpenAI and Anthropic have recently launched medical conversations: ChatGPT Health and Claude for Healthcare, allowing users to connect their personal medical records to the AI platforms. Concerns remain, however, regarding patients’ increasing reliance on AI due to healthcare systems being perceived as inaccessible or dismissive. This phenomenon is referred to as a “watershed moment,” collapsing the traditional information asymmetry where physicians held the data and patients held only the symptoms.

OpenAI has reported that over 230 million people ask ChatGPT health-related questions each week, typically driven by the inaccessibility of modern healthcare.4 This data was acknowledged by the development of ChatGPT Health, a feature that allows users to center medical conversations on their personal health records. In a matter of days, Anthropic followed the movement by releasing Claude for Healthcare, offering similar functionality while recommending that qualified professionals review AI-generated outputs before making a final decision. 

These developments have prompted discussion regarding whether these tools offer powerful personalization of patient symptoms and medical records ready for clinical use, and if they pose risks in the form of limited regulatory oversight and the potential for “hallucinations.” Hallucinations refer to AI-generated responses that are presented confidently despite lacking factual basis.

The article references several studies performing a comparative analysis evaluating the efficacy and safety of artificial intelligence in healthcare contexts. A 2023 cross-sectional study published in JAMA Internal Medicine compared physician responses with ChatGPT’s responses to patient inquiries. Researchers found that licensed healthcare professionals preferred the AI-generated responses approximately 79% of the time because they were perceived as more empathetic and comprehensive.1 However, several important limitations remain. A study by the Icahn School of Medicine at Mount Sinai revealed that AI chatbots provided incorrect information in 50–82% of the cases when presented with false medical information, often elaborating on the misinformation.3 Additionally, when responding to conversations involving mental health concerns, OpenAI has consulted with over 170 mental health professionals to improve ChatGPT’s responses to conversations involving emotional distress and poor mental well-being4. According to OpenAI, these updates improved the model’s responses in self-harm-related conversations by 52%.4 

OpenAI has stated that most health-related conversations occur outside of normal clinic hours, suggesting that AI is filling in a critical informational gap for individuals considering whether to seek emergency medical care.

These developments may reshape healthcare provider and patient interactions by allowing physicians to discuss AI-generated information without dismissing patients’ concerns, while simultaneously providing professional insight that helps patients better understand their condition. Physicians may benefit from integrating AI for administrative tasks such as synthesizing clinical data. Healthcare organizations have also been encouraged to establish “AI formularies,” curated lists of approved medical AI tools with established guidelines while broader federal regulations continue to evolve. Researchers have also noted that, even though autogenerated texts can provide explanations, they cannot be held accountable, highlighting the need for legislative reform. Future legal frameworks must clarify whether liability lies with developers, physicians, or both when AI provides inaccurate medical advice. Furthermore, greater transparency regarding data privacy is also needed, as consumer AI products may operate outside the protections provided by HIPAA, the Health Insurance Portability and Accountability Act designed by the federal government to establish national standards for electronic health records.

References

  1. Ayers, J. W. et al. Comparing physician and artificial intelligence chatbot responses to patient questions posted to a public social media forum. JAMA Intern. Med. 183, 589–596 (2023).
  2. Bascom, E. Integrating medical records with AI ‘a watershed moment’ in medicine – is it dangerous? Healio Gastroenterol. (2026)
  3. Mount Sinai Health System. AI chatbots can run with medical misinformation, study finds, highlighting the need for stronger safeguards. Mount Sinai Newsroom. (2025) 
  4. OpenAI. Introducing ChatGPT Health. OpenAI

Related Posts

Darius Dastur,  Mishal Ali, Harshitha Ezhilarasan, Hasan Khan, Aaron Akasala

Add Your Heading Text Here