On June 18, 2026, OpenAI announced that it had improved the quality of ChatGPT's responses in the area of health[1]. The newly available GPT‑5.5 Instant, now offered to all free users, shows improvements in recognizing when urgent care may be needed and in explaining uncertainty, and OpenAI says it now performs on par with the company's higher-end "Thinking" models on its most difficult health evaluations[1].
230 Million People a Week Ask Health Questions, and the Gains Reach Free Users
According to OpenAI, health and wellness questions are one of the main ways people use ChatGPT, with more than 230 million people each week turning to it to make sense of lab results, prepare for appointments, navigate insurance, and build healthier habits[1].
With GPT‑5.5 Instant, OpenAI reports gains in recognizing when urgent care may be needed, asking for relevant context, explaining uncertainty, and making complex information easier to understand[1]. What stands out is that these improvements are not limited to paid plans but are available to every free ChatGPT user[1]. The aim is to raise the quality of responses across a wide range of situations, from everyday wellness questions to more complex clinical ones.
A Review System Backed by 260-Plus Physicians Across 60 Countries
Behind the quality gains is a physician-led evaluation process. OpenAI says it works with a network of more than 260 physicians across 60 countries, 49 languages, and 26 medical specialties to define what a "good" answer looks like in real-world health situations[1].
The evaluations rely on the company's openly available "HealthBench" and the more specialized "HealthBench Professional"[1]. HealthBench is an evaluation standard OpenAI released that uses realistic health conversations and physician-written rubrics to measure qualities such as accuracy, safety, clarity, context awareness, and completeness[2]. Physicians have reviewed more than 700,000 example model responses to date, with a new response checked every few minutes[1].
Compared With Physician Answers, and Fewer Signs of Error
OpenAI also ran a test in which physicians wrote responses with no limits on time or sources (but without using AI), and a separate panel of physicians compared those with the model's responses[1]. In a comparison of 3,500 responses, GPT‑5.5 Instant showed fewer failure modes — such as not adapting to local healthcare context, missing red flags or referrals to care, and failing to ask for needed information — than both older models and the physicians' own responses[1].
Real-world usage points the same way. Using privacy-preserving monitors on production traffic, OpenAI compared billions of messages a week and reports that the share of responses with at least one suspected factuality issue fell by 71 percent over the past two months[1].
Expanding Into Tools for Medical Settings
OpenAI says this work also feeds into tools designed for medical professionals, such as "ChatGPT for Clinicians" and "OpenAI for Healthcare," which support documentation, research, and care delivery[1]. The company frames improving human health as "one of the most personal, tangible impacts of AGI" and says it plans to bring these gains to more people[1].
That said, all of the figures here come from OpenAI's own evaluations and tallies. ChatGPT is meant to help organize information and support questions, not to replace decisions about diagnosis or treatment. If you have symptoms that concern you, it is important to see a medical professional.
Summary
OpenAI announced that GPT‑5.5 Instant, available even to free users, has improved health-related responses to a level comparable with its top models. With an evaluation system involving more than 260 physicians across 60 countries, and data showing a 71 percent drop over two months in responses flagged for possible errors, ChatGPT's health guidance is moving in a safer and more practical direction. At the same time, the published figures rest mainly on the company's own testing, so leaving final health decisions to professionals remains essential.
Source: https://openai.com/index/improving-health-intelligence-in-chatgpt/
