Artificial intelligence is being pitched as a fix for almost every bottleneck in healthcare, but at least one stubborn problem is resisting the sales pitch: translating prescription information for patients who don't read English.
According to Healthcare IT News, in a piece headlined "AI alone cannot solve Rx translation," automated language tools are not sufficient on their own to handle the job. The framing is notable because prescription translation looks, on the surface, like an ideal machine-translation task — short, repetitive, highly standardized text. The publication's conclusion is that technology by itself doesn't close the gap.
The stakes here are unusually concrete. A mistranslated dosage instruction isn't a garbled sentence in a chatbot window; it's a patient taking a drug twice a day instead of every other day. Pharmacy labels compress critical instructions into a few words, and small errors carry outsized consequences. That is precisely the kind of context where the confident-sounding output of a language model can be more dangerous than an obvious failure, because nothing about the result looks wrong to the person reading it.
Beyond the specifics of the pharmacy counter, the story fits a broader pattern in health technology: AI tools tend to work best as assistants inside a workflow that still includes human review, rather than as drop-in replacements for it.
It matters because the industries rushing hardest to automate language work are often the ones where a quiet translation error does the most damage.