Key Highlights
- ChatGPT attracts over 230 million weekly users seeking health and wellness guidance
- Former Google Med-PaLM architect Karan Singhal now spearheads OpenAI’s medical AI initiatives
- GPT-5.5 Instant demonstrated superior performance compared to physician-generated responses in evaluation metrics
- Medical misinformation incidents decreased 71% over a two-month period in live deployment
- The ChatGPT Health platform, introduced in January, integrates with wellness applications and electronic health records despite ongoing waitlist
OpenAI reports that ChatGPT now serves more than 230 million weekly users seeking guidance on health and wellness topics. These interactions range from interpreting laboratory test results to planning upcoming medical consultations.
The organization has positioned healthcare as a strategic priority, with researcher Karan Singhal directing this initiative.
Karan Singhal’s Background and Role
Singhal transitioned to OpenAI during mid-2024 following his tenure at Google, where he played a pivotal role in developing Med-PaLM, a suite of artificial intelligence models tailored for healthcare queries. Google subsequently reduced funding for this specialized project, pivoting toward broader-purpose AI systems.
Upon joining OpenAI, Singhal established a dedicated healthcare research division and collaborated with over 200 medical professionals. This partnership aimed to integrate authentic clinical expertise into ChatGPT’s health-related responses.
He spearheaded the development of HealthBench, a specialized assessment framework created in collaboration with this physician network to evaluate and benchmark AI medical performance.
GPT-5.5 Instant Performance Metrics
OpenAI’s newest publicly available model, GPT-5.5 Instant, underwent rigorous testing against both physician-authored responses and previous AI iterations. The model achieved superior scores in multiple categories including precision, communication quality, thoroughness, and utility for healthcare decisions.
Medical professionals evaluating the system noted significantly reduced failure patterns. These potential issues encompass overlooking critical warning signs, insufficient contextual questioning, or neglecting to recommend professional medical consultation when appropriate.
According to OpenAI’s analysis of billions of weekly health-related conversations, instances of flagged factual inaccuracies declined by 71% during the preceding two-month period.
Notably, the GPT-5 series represents OpenAI’s first model family engineered with healthcare optimization integrated throughout the entire training process.
Enhanced Patient Context Understanding
A fundamental limitation of AI healthcare assistants involves their lack of user-specific information. Traditional physicians access comprehensive medical histories and records, while chatbots typically operate without such background data.
To address this gap, OpenAI introduced a specialized health feature within ChatGPT in January, enabling integration with wellness applications and allowing medical record uploads. Singhal illustrated this capability by describing how he uploaded sleep tracking data from his Apple Watch, which the system analyzed to determine that elevated bedroom temperature was compromising his deep sleep quality.
Despite launching over five months ago, ChatGPT Health continues to maintain a waitlist for new users.
Singhal emphasized the team’s commitment to enhancing the chatbot’s diagnostic questioning approach, mirroring the methodical inquiry process physicians employ before providing recommendations. He noted that medical professionals are rapidly adopting AI tools, with institutional resistance proving less substantial than initially anticipated.
Ongoing Challenges and Expansion
OpenAI currently faces active litigation claiming that an earlier iteration, GPT-4o, provided potentially dangerous medical guidance to users. The company has contested these allegations, with legal proceedings still underway.
Neverthstanding these legal challenges, OpenAI has maintained its commitment to expanding healthcare functionalities rather than retreating from the sector.
GPT-5.5 Instant remains accessible to all ChatGPT users without requiring a premium subscription, democratizing access to enhanced medical response capabilities.


