OpenAI just launched ChatGPT Health, a dedicated space inside ChatGPT for health conversations on mobile and web. The primary feature is simple but necessary: you can securely connect medical records and wellness apps like Apple Health, Function Health, and Peloton. This allows the assistant to respond with your specific history instead of generic advice. While some argue whether consumers should ask AI about health, the reality is that the demand is already here. OpenAI reports that over 40 million people use ChatGPT daily for health questions, with roughly 230 million weekly queries. This happens because the existing system fails at the basics of time, continuity, and clarity. Some users have even started taking this into their own hands elsewhere, such as using Claude Code to analyze their own DNA data.
Stats cited in launch materials show a system under heavy strain, driving users toward AI tools that offer immediate feedback.
The context gap in modern medicine
Fidji Simo shared a personal story that explains why this exists. While hospitalized, a resident prescribed an antibiotic that could have triggered a dangerous complication based on Simo’s specific history. The patient caught it because she had uploaded her records to ChatGPT. The resident was not defensive; she was relieved. She explained that she only had five minutes per patient and the records were not organized to make such risks obvious. This is the core problem AI helps solve. It is not about replacing medical judgment, but about surfacing context that is already in the record but hidden by administrative friction and time pressure. In my 2026 AI predictions, I noted that agents get real, and this is a prime example of an agent handling the cognitive load of data synthesis.
A vertical product approach
Today, users paste PDFs or lab results into a chat and hope for the best. ChatGPT Health turns this into a formal workflow. It is a private space where you connect sources directly. This includes secure connections to medical records, b.well integration for financial and wearable data, and Function Health for lab biomarkers. Initially, EHR connections are U.S.-only, starting with waitlist users and expanding to iOS and web. This is a smart wrapper. The model is the engine, but the wrapper determines if the model is usable for a real problem. In healthcare, the wrapper is the product because it handles the messy pipes of data integration.
Addressing structural failures
The healthcare system suffers from four main issues. First, clinician bandwidth is gone. When doctors have minutes, reviewing a full chart is impossible. AI can pre-summarize risks and medication interactions to reduce mistakes. Second, records are fragmented. Only a small percentage of physicians fully exchange electronic data. ChatGPT Health makes the patient-side view coherent even when the institutional view is not. Third, access and cost are barriers. Many AI health conversations happen outside clinic hours because the system is unavailable. Finally, healthcare is reactive. Most outcomes are driven by daily routines like sleep and diet. A daily assistant that interprets wearable trends from Apple Health or Peloton is a step toward prevention that the current billable system cannot provide. This mirrors the trend I discussed in AI chatbot market share, where existing attention makes vertical expansion easier.
How to use it effectively
Treating this like a doctor is a mistake. Treating it like a high-end organizer is effective. It is useful for appointment prep, translating lab results into plain language, cross-checking medication history, and habit support. It is not for emergency triage or replacing a clinician for treatment changes. Privacy remains the main bottleneck. OpenAI is excluding health data from training by default and requiring MFA. The utility of the tool depends on whether OpenAI maintains these strict defaults as the product scales. If you want to see how this fits into the broader 2026 ecosystem, check out my notes on quality versus cost in the current model market.
