A vet read a blood panel as 2.8% RBC (red blood cell count). The real number was 22.8%. One misplaced decimal point triggered a recommendation to euthanize a cat that was still jumping onto furniture and eating normally.
The owner, whose cat has chronic kidney disease, took the results to ChatGPT before acting. The AI flagged an immediate problem: a 2.8% RBC count is physiologically incompatible with a cat displaying normal activity. At that level, the animal would be in acute crisis - barely able to stand, let alone move around. ChatGPT told her the numbers didn't add up and suggested getting a retest.
The retest came back at 22.8%. The vet hospital had made a transcription error. The cat is alive.
What the AI Actually Did
ChatGPT didn't diagnose the cat. It did something more useful in that moment: it cross-referenced the reported numbers against what it knows about feline biology and identified an inconsistency. That's a sanity check, not a diagnosis.
This is where large language models - AI systems trained on vast amounts of text, including medical and veterinary literature - can perform a specific and useful function. They can flag when a reported value clashes with expected physiological ranges. They can't run labs or examine the animal, but they can catch obvious numerical impossibilities.
The error here was a classic transcription mistake: one digit moved one decimal place. These happen in medical settings with more frequency than patients realize. Lab software, manual data entry, phone-based reporting - any handoff point is a potential failure point. A 10x discrepancy in a critical value is exactly the kind of error that should trigger an automatic check, and in this case, it didn't.
The Second Opinion You Can Always Get
For people who can't easily get a second professional opinion on short notice, AI has become a fast pressure-test for alarming results. You paste in the numbers, describe what you're observing, and ask: "Does this make sense?"
That's not a replacement for medical expertise. ChatGPT can be wrong about medical details - it can misremember specific thresholds or get drug interactions wrong. Any AI-flagged concern still needs verification from an actual professional.
But "does this make sense at all?" is a different question from "what should I do?" The first is a logic check. This case shows AI can handle that well when the error is as stark as a tenfold discrepancy between reported value and observable reality.
The practical takeaway isn't "trust AI over your vet." It's closer to: you're allowed to push back on results that don't match what you're seeing. You're allowed to ask for a retest. ChatGPT gave this owner a specific, articulable reason to question the numbers rather than just a gut feeling - and that was enough.