AI Wants Your Medical Records Now

Three major AI platforms have launched health products in ten weeks. OpenAI went first with ChatGPT Health in January, Microsoft followed with Copilot Health on March 12, and Perplexity closed the sprint on March 20 with Perplexity Health, a suite of data connectors pulling Apple Health, Fitbit, Withings, Ultrahuman, and electronic health records into one interface. The pitch is personalization: answers grounded in your actual data rather than generic population averages. The timing across all three companies is too tight to be anything other than a coordinated market response to the same opportunity.

Your Health Data Is Already a Disaster

The problem Perplexity is solving exists independent of any competitive motive. Health data is spread across a hospital portal you log into twice a year, lab PDFs buried in your email, a wearable syncing to its own app, and a pharmacy dashboard with a different login entirely. None of these systems share information with each other, which means every doctor's appointment starts with an incomplete picture and every patient has to reconstruct their own history from memory.

Perplexity's infrastructure partner here is b.well Connected Health, a HIPAA-compliant platform connected to over 2.4 million providers and more than 350 health plans and labs across the US. The practical result is that a question about resting heart rate can draw on recent activity data, cardiac history, and latest bloodwork simultaneously rather than requiring the user to manually pull and describe all three.

The Privacy Pledges

Perplexity says health data is encrypted in transit and at rest, will not train their models, will not be sold, and can be fully deleted by the user at any time. Microsoft and OpenAI made nearly identical statements when they launched their respective health products. Independent audits of any of these commitments at meaningful scale have not happened yet, which is simply a function of how recently these products launched rather than evidence of bad faith. Whether that scrutiny materializes, and what it finds, will matter more to long-term adoption than anything in a launch announcement.

The Liability Disclaimer That Does Real Work

Perplexity, like its competitors, is explicit that this product is not a diagnostic tool. The framing positions it as preparation for clinical conversations: a way to walk into an appointment with organized data and informed questions rather than a replacement for the appointment itself. Legally that framing is sound. Practically it starts to strain when someone is searching symptoms at midnight with their full health history connected and gets a confident, cited, personalized-sounding answer. A Washington Post investigation earlier this year found ChatGPT returning health information not supported by the data it was given, a problem that sits in the model itself rather than the data pipeline, and one that a Health Advisory Board of physicians and researchers can pressure-test but not eliminate.

The Actual Business Logic

Perplexity Health is the second vertical built on Perplexity Computer, following Perplexity Finance which used Plaid to connect brokerage accounts. Competing with OpenAI on raw model capability is a losing proposition for a company Perplexity's size. Owning specific verticals where deep data integration makes the product genuinely different is a more defensible position. An AI connected to your health records and your financial accounts, with your consent and real infrastructure behind both integrations, is harder to replicate and harder to leave than a general-purpose chatbot. User willingness to actually hand over that data is the variable that determines whether any of this matters.