The AI Health Wars Have Begun
In the span of ninety days, three of the world's most powerful artificial intelligence companies have launched competing products designed to do something that would have seemed reckless just a year ago: connect directly to your medical records, fitness trackers, and lab results, then use that data to answer your health questions.
Perplexity fired the latest shot on Thursday, unveiling Perplexity Health, a suite of data connectors that pulls together electronic health records from more than 1.7 million care providers, wearable data from Apple Health, Fitbit, Ultrahuman, and Withings, and lab results — all funnelled into a single AI-powered interface that promises personalised health answers grounded in your actual medical history rather than generic population data.
The product, built on top of the company's Perplexity Computer agent platform, is rolling out to Pro and Max subscribers in the United States. It follows Perplexity Finance as the second major vertical the company has constructed around its autonomous agent infrastructure, signalling a deliberate strategy to move beyond search and into deeply personal data domains.
Perplexity is not pioneering this space so much as sprinting to catch up. OpenAI launched ChatGPT Health in January, integrating with Apple Health to let users ask questions informed by their fitness and wellness data. Microsoft followed on March 12 with Copilot Health. Amazon and Anthropic have also debuted healthcare-focused AI products this year. What was an empty category twelve months ago is now one of the most fiercely contested territories in the industry.
The architecture behind Perplexity's offering is worth examining. Electronic health records are pulled through b.well Connected Health, a HIPAA-compliant platform whose network spans more than 2.4 million providers and 350 health plans and labs across the United States. Wearable data flows through Terra API, a unified health and fitness data aggregator. The result is a system that can, in theory, cross-reference your resting heart rate trends with your cardiac history and most recent bloodwork simultaneously — the kind of synthesis that typically requires a physician with access to your complete file.
The company is positioning the product for tasks like generating pre-appointment visit summaries, personalised nutrition plans, and training protocols, with responses drawing from clinical guidelines and peer-reviewed journals. A newly formed Health Advisory Board of physicians and researchers will pressure-test product decisions against evidence-based medicine standards.
Every company in this race is making the same privacy pledges: health data encrypted in transit and at rest, never used to train AI models, never sold to third parties. Every company is also attaching the same disclaimer: this is not a diagnostic tool, and it does not replace professional medical advice.
But the gap between how these products are marketed and how they will inevitably be used is the central tension of this emerging category. A Washington Post investigation earlier this year found that ChatGPT was liable to report health information not supported by the data it was given — a baseline reliability problem that no amount of data connectivity resolves if the underlying model hallucinates. When the hallucination concerns your cardiac history rather than a trivia question, the stakes shift dramatically.
The speed of this race also raises questions about regulatory readiness. The FDA's framework for AI in healthcare was designed primarily for clinical decision-support tools used by physicians, not consumer-facing products that millions of people might use to interpret their own lab results at two in the morning. No major regulatory body has issued specific guidance on AI products that aggregate personal health records for consumer Q&A purposes.
For the companies involved, the strategic logic is clear. Health data represents perhaps the ultimate lock-in: once a user has connected their medical records, fitness trackers, and lab portals to a single AI platform, switching costs become enormous. The company that wins this category does not just gain a user — it gains the most intimate dataset a person possesses.
Kristen Valdes, founder and CEO of b.well, framed the rationale simply: AI health questions are already happening at scale. The question is whether the answers are grounded in a person's actual medical history or in generic data scraped from the open web. That framing is compelling, but it papers over an uncomfortable truth. The technology industry has a long history of identifying genuine problems, building solutions that create new and different problems, and then asking for trust while the consequences sort themselves out.
The AI health wars have begun in earnest. Whether patients end up better informed or merely better marketed to will depend on details that no press release can resolve.
