OpenAI's ChatGPT Health: A Game-Changer for Patient Care?

Health and wellness already rank among the most common reasons people use AI chatbots like ChatGPT. 

In fact, according to OpenAI, more than 230 million users worldwide ask health-related questions every week.

Perhaps it was the natural progression, then, for OpenAI to launch ChatGPT Health, which brings together personal health information and AI support. The new feature is said to build on existing security and privacy features by introducing further protections specific to sensitive health information, while keeping the experience accessible and useful.

OpenAI has been keen to emphasise that, instead of replacing professional medical advice, ChatGPT Health is designed to work alongside it. The service has been designed in collaboration with physicians and offers encrypted, isolated chats where users can link their health data from medical records or wellness apps.

These features give users relevant, contextual support for day-to-day health questions, diet and fitness planning, understanding test results or even navigating insurance options.

Meanwhile, OpenAI rival firm Anthropic has just launched Claude for Healthcare, highlighting growing competition among AI specialists to gain a foothold in healthtech. 

A separate space designed for health

Healthcare information often exists in separate portals, apps, wearables or scattered documents, making it difficult to understand the full picture.

ChatGPT Health is billed as a single space where medical data, wellness app insights and health chats live together but are kept separate from general ChatGPT conversations.

The separation is very much deliberate. Health conversations are stored with their own dedicated memories, meaning they do not mix with general chat history. If a health-related topic comes up in a standard chat, ChatGPT may prompt a move to the Health section for stronger protections.

These health chats remain visible in the user’s conversation history, but the content stays isolated.

Users can also bring in external health information by connecting platforms such as Apple Health, MyFitnessPal or Function. In the US, integration with electronic medical records is made possible through b.well, a secure network that links patients to a broad range of healthcare providers. b.well meets high industry standards for privacy and security, and users can revoke access to records or apps at any time through the Settings menu.

When helpful, ChatGPT Health may pull limited context from non-health conversations – such as a mention of a lifestyle change – but health data and memories will never flow out into general chat. OpenAI says files, conversations and memory entries inside the Health space will remain locked there unless deleted or edited directly by the user.

Built with data protections and physician input

ChatGPT Health is built to use the same baseline security applied to ChatGPT overall, including encrypted storage and transmission of data. Due to the nature of health information, it adds layers of protection, including custom-built encryption and system isolation.

OpenAI has highlighted that health-related chats are never used to train its foundation models.

Users also have control over what gets saved or shared. Options like temporary chats and memory deletion are available across the platform, while multi-factor authentication (MFA) adds another level of control to guard against unauthorised access.

When it comes to third-party health apps, all those integrated into ChatGPT Health must meet strict security and privacy criteria. They must go through an additional security review before inclusion and may only collect the minimum amount of data required. Users receive clear information about the type of data each app accesses during the connection process and can remove apps from their profile at any time.

The product itself has been shaped through consultation with medical professionals. Over the past two years, OpenAI says it has worked with more than 260 physicians across 60 countries and a broad range of specialisations.

The group has reviewed model responses more than 600,000 times using physician-developed scoring methods, guiding how the model responds to queries, signals urgency and communicates safety-related concerns.

ChatGPT Health uses a testing system called HealthBench, created with physician input, to measure its responses against clinical standards. It focuses on quality as judged by clinicians in real-life situations, with factors including whether responses are clear, escalate care where appropriate and provide context-sensitive advice without assuming diagnosis.

Industry reaction

ChatGPT Health has, so far, garnered mixed reaction from healthcare and technology experts alike. 

Max Jones, Chief AI Officer at RPNA and a Non-Executive Director in the UK's NHS, emphasised that the platform "comes with real risks", but said the public had to think carefully about decisions concerning their health. 

Writing on LinkedIn, he added: "The power rests not with clinicians, legislators or regulators; rather, the power is in the hands of the public through the decisions they take when confronted with a health uncertainty. Should they wait for authoritative human care or seek more immediate guidance from an LLM?

"Until the NHS recovers its service performance and public confidence, I suspect we will see the public drive LLM-based triage whether the professionals like it or not."

Speaking to the BBC, Max Sinclair, Founder and CEO of AI marketing platform Azoma, called the release of ChatGPT Health a "watershed moment" in the context of agentic commerce. 

Taking to LinkedIn, he continued: "You would trust this more than a 15-minute assessment from an overworked GP who has been back-to-back all day before seeing you. Comparative diagnostic accuracy studies between ChatGPT and humans would prove you right to do so."

Ben Legg, Co-Founder and CEO of The Portfolio Collective, comments: "We are moving from using AI as a secretive second opinion to a primary triage tool. In the future, before you see a human, an agent will likely review your medical history, analyse your wearables data and summarise your symptoms for the consultant.

"This offers incredible efficiency and could solve the GP appointment backlog crisis overnight. But it fundamentally changes the patient-doctor relationship."

ChatGPT Health is initially being rolled out to a limited number of users with Free, Go, Plus or Pro plans, excluding those in the European Economic Area, Switzerland and the UK. Medical record and app integration features are currently available only in the US, while Apple Health integration requires iOS.