OpenAI improves ChatGPT on health questions

OpenAI has made a health-focused upgrade to ChatGPT that matters because health is already one of the most common ways people use the product. In a June 18, 2026 product update, the company said GPT-5.5 Instant now gives substantially better answers to health and wellness questions, and that the model is available to free ChatGPT users, subject to normal limits.

The scale is the first reason this is important. OpenAI says more than 230 million people use ChatGPT each week for health-related questions, from understanding lab results and preparing for appointments to making sense of symptoms, insurance and daily wellness choices. That does not turn ChatGPT into a doctor, but it does mean small improvements in clarity, caution and factuality can affect a very large audience.

According to OpenAI, GPT-5.5 Instant performs at a level comparable to its frontier Thinking models on challenging health evaluations. The company says the new model is better at spotting situations where urgent care may be needed, asking for relevant context, explaining uncertainty and making complex information easier to understand. It also says privacy-preserving monitoring of production traffic shows a 71 percent drop in health responses with at least one flagged factuality issue over the last two months.

The work behind the update is also notable. OpenAI says more than 260 physicians across 60 countries, 49 languages and 26 medical specialties helped define and review what good health responses should look like. Physicians have reviewed more than 700,000 example model responses, creating rubrics for safety, accuracy, communication and appropriate escalation.

For AI users, the message is practical: health answers should become more useful, but still require judgment and professional care. For makers and companies, the update shows where consumer AI is heading: not just more powerful models, but domain-specific evaluation, expert review and risk monitoring. In sensitive fields, the product story is increasingly about reliability, not only raw intelligence.