Somewhere in the time it takes a hospital committee to schedule a meeting about whether AI is even allowed in the building, actual patients have already opened a laptop, pasted in their lab results, and gotten an answer. It's less a controlled rollout and more everyone quietly walking past the manned register and scanning their own groceries while management is still arguing about the signage.

Large language models like GPT-5 and Claude are genuinely good at pattern-matching across millions of pages of medical literature. Ask one what a creatinine level of 1.8 means for a 55-year-old man, and it'll give a coherent, often useful answer, sometimes a better one than a rushed five-minute appointment allows for. The catch is it also hallucinates with total confidence, doesn't know your full medical history, and has never actually examined you, three fairly important asterisks buried under all that fluent, reassuring prose.

Roughly a third of adults now turn to AI chatbots for health advice, and about one in five use it specifically to interpret lab results, according to a 2026 KFF poll. OpenAI says more than 300 million people ask it health-related questions every week. Meanwhile, an MIT study found the part worth actually sitting with: non-experts tended to trust the AI's answer even when it was flat-out wrong, while the clinicians in the same study were the ones who caught its mistakes.

The most regulated industry on the planet is being lapped by a twenty-dollar-a-month subscription, largely because nobody's finished arguing about the rulebook yet. UnitedHealth thinks AI could save it close to a billion dollars this year. The governance memo, presumably, is still in committee.

This is WebMD's plot twist: the exact self-diagnosis instinct everyone made fun of for a decade, except now it's genuinely competent, and still occasionally tells you it's cancer when it's a cold.

Sources: WSJ (via ToVima) — Doctors Don't Want Patients to Read Test Results With AI. They're Doing It Anyway. (https://www.tovima.com/wsj/doctors-dont-want-patients-to-read-test-results-with-ai-theyre-doing-it-anyway); NPR (via NCPR) — Running Your Lab Results by ChatGPT? Here's What to Keep in Mind (https://www.northcountrypublicradio.org/news/npr/nx-s1-5537067/running-your-lab-results-by-chatgpt-here-s-what-to-keep-in-mind)