Meet the top OpenAI researcher trying to make ChatGPT into a Google-sized healthcare juggernaut
Full article text is available in the Catalayer news terminal.
Summary
OpenAI researcher Karan Singhal, who leads the company's health initiative, says more than 230 million people use ChatGPT for health and wellness advice each week, and that GPT-5 is the company's first model trained specifically for health quality at every stage of development. The latest free model, GPT-5.5 Instant, outperformed both physician-written answers and GPT-4o in internal evaluations using OpenAI's HealthBench testing framework.
Market Impact
OpenAI's push into health represents a direct competitive challenge to Google Search's dominant position in online health information, with Singhal — who previously built Google's Med-PaLM models — framing conversational AI as a structural upgrade over static web search. The 71% reduction in flagged inaccurate health responses over two months reflects a rapid quality iteration cycle enabled by HealthBench, a physician-guided evaluation framework co-developed with more than 200 clinicians. This analysis is informational and avoids any directional trading claims.
Why It Matters
It documents the most detailed public account of OpenAI's healthcare strategy, including the scale of weekly health use and the company's approach to clinical validation, just as AI health products begin to face regulatory and liability scrutiny.
Key Points
- More than 230 million people use ChatGPT for health and wellness advice each week, according to OpenAI; the company's GPT-5 model family is the first trained for health quality at every stage of development.
- OpenAI's HealthBench testing framework was developed with more than 200 physicians; GPT-5.5 Instant scored better than physician-written answers and GPT-4o in HealthBench tests.
- The company recorded a 71% drop in health responses flagged for inaccuracy over two months; OpenAI's January-launched ChatGPT Health product, which connects to health apps and medical records, still carries a waitlist more than five months after launch.
- Karan Singhal, who previously built Google's Med-PaLM medical AI models at Google before the firm cut that investment in favor of general-purpose models, is leading the OpenAI health research team.
Key Entities
Evidence
More than 230 million people use the tool for health and wellness advice each week, according to OpenAI.Supports: Supports the weekly user count.
OpenAI's GPT-5 model family is the company's first to be trained specifically at every stage of development to be better at health advice.Supports: Supports the model-training claim.
OpenAI's latest free model, GPT-5.5 Instant, scored better than both physician-written answers and GPT-4o in tests, the company said Thursday. Comparing billions of anonymized messages about health, they also said the...Supports: Supports the benchmark and improvement figures.