It's 10pm on a Friday and a lab portal notification just landed: a flagged result, a term you don't recognize, a reference range you can't parse. The office closed hours ago. The earliest appointment is two weeks out. Googling the term surfaces a mix of a Mayo Clinic page, a decade-old forum thread, and a supplement ad. This exact moment, confusing information, no clinician available, real anxiety, is the gap every major AI health platform is racing to fill. They are not really competing with each other. They're competing with that broken chain: search, forum, wait, forget your questions, sit across from a doctor still confused.

Five Companies, Five Different Bets

  • OpenAI Health is positioned as the explainer: plain-language answers to "what does this mean," strong at translating lab values, discharge summaries, and medical jargon into something a layperson can actually use.
  • Microsoft Copilot Health is positioned as the navigator: it leans on Microsoft's enterprise footprint in hospital systems and electronic health records to help people move through the healthcare system itself, finding the right specialist, understanding a referral, organizing records across providers.
  • Amazon Health AI is positioned as the operator: closer to logistics than explanation, prescription management, appointment scheduling, pharmacy integration, the operational plumbing of getting care done.
  • Perplexity Health is positioned as the research assistant: citation-first answers built around retrieval-augmented generation, showing its sources rather than asserting an answer, which matters directly for the hallucination risk covered in our companion piece below.
  • Verily blends AI and clinician judgment directly, routing AI-generated output through actual clinical review rather than shipping the model's answer straight to the patient, a slower but more conservative model.

What They All Do Well

Across all five positions, the common strength is friction reduction, specifically the friction that happens before and after a doctor visit, not during it. That includes explaining medical information in plain language, organizing scattered health records into something coherent, and generating a sharper list of questions to bring into an appointment instead of showing up with a vague sense of worry. None of that requires clinical judgment. It requires patience and clarity, which is exactly where these tools are strongest.

What None of Them Can Do Yet

  • Diagnose. Every platform in this category positions itself, explicitly or through its terms of service, as informational rather than diagnostic. Pattern-matching symptoms to a plausible condition is not the same as a clinical diagnosis.
  • Prescribe. None of these tools initiate or adjust medication independently of a licensed prescriber.
  • Account for a patient's full history. Even the most integrated platforms work from whatever records happen to be connected, not the complete, longitudinal picture a long-term treating physician has.
  • Replace clinical judgment. Judgment calls that weigh a patient's specific context, risk tolerance, and full picture against textbook guidance remain a human function, and every platform here is built around that boundary rather than against it.

What They Actually Compete On

Strip away the marketing and the real competitive dimensions are: how the data privacy model works (on-device processing versus cloud, and how deeply it integrates with existing electronic health records), the depth and currency of the underlying medical knowledge base, and hallucination rate, how often the system states something false with total confidence. That last dimension is significant enough that it deserves its own treatment; see our full breakdown of what the research shows about AI hallucinations in healthcare.

Where This Is Headed in the Next 3-5 Years

The clearest trajectory is integration with wearables and continuous monitoring, moving from a one-time chat interaction toward an ongoing signal: a platform that notices a trend in resting heart rate, sleep quality, or glucose variability before a person would think to ask about it. That connects directly to the biomarker and wearable tracking ground covered elsewhere on this site, and to the broader question of how AI is already being used in longevity research to identify which biomarker signals actually predict outcomes. Earlier detection of a meaningful trend, flagged early enough to matter, is the realistic next frontier, not autonomous diagnosis. For the practical, day-to-day version of this question, Age Better Today's guide to AI tools for healthy aging covers how people are already using these tools in daily life.

Important Caveat

AI health platforms are informational tools, not a substitute for a licensed healthcare professional. Nothing in this article, or in any of the platforms discussed, should be treated as a diagnosis, a prescription, or a reason to delay care for a genuine medical concern.