Hospitals Are Deploying AI Like Crazy. We Still Don’t Know If It Helps Patients

I don’t need to tell you that AI is everywhere. It’s in your phone, your car, your search engine, and increasingly, your doctor’s office.

Doctors are using AI to take notes during appointments. Algorithms are crawling through patient records, flagging people who might need extra support or specific treatments. Radiology departments are leaning on AI to interpret X-rays and scans. The hype cycle has hit healthcare hard, and adoption is accelerating fast.

Here’s the uncomfortable truth: a growing pile of studies shows many of these tools can deliver accurate results. But accuracy isn’t the same as usefulness. The bigger question — the one that actually matters — is whether using these tools translates into better health outcomes for patients.

We don’t have a good answer. And that’s precisely the point Jenna Wiens and Anna Goldenberg make in a paper published this week in Nature Medicine.

Wiens, a computer scientist at the University of Michigan, has spent years trying to get clinicians interested in AI. For the first decade of her career, she says, it was an uphill battle. Then, over the last few years, something flipped. Healthcare providers aren’t just curious anymore — they’re deploying these tools at scale, often without rigorous evaluation.

Take “ambient AI” scribes. These tools listen to conversations between doctors and patients, then transcribe and summarize them. Multiple vendors offer them, and adoption has been rapid. A few months ago, a staffer at a major New York medical center told me that doctors are “overjoyed” — the tech lets them focus entirely on the patient during appointments and eliminates hours of paperwork. Early studies back up the anecdotal evidence, suggesting scribes reduce clinician burnout.

All of that is fine. But what about patient health? “[Researchers] have evaluated provider or clinician and patient satisfaction, but not really how these tools are affecting clinical decision-making,” Wiens told me. “We just don’t know.”

The same problem applies across the board. Some AI tools predict a patient’s health trajectory. Others recommend treatments. They’re all designed to make healthcare more effective and efficient. But even a tool that’s technically “accurate” doesn’t guarantee better outcomes.

Consider chest X-rays. AI can speed up interpretation significantly. But how much will a doctor actually rely on that analysis? How does the tool change the way they interact with the patient or recommend treatment? And ultimately, what does that mean for the person lying on the exam table? The answers almost certainly vary between hospitals, departments, and even individual doctors at different stages of their careers.

Wiens raises another concern with the scribes. Research on AI in education suggests that offloading certain cognitive tasks can change how people process information. Could the same thing happen with doctors? If a medical student or resident relies on an AI summary instead of forming their own mental model of a patient’s case, does that affect their clinical judgment down the line? “We like things that save us time, but we have to think about the unintended consequences of this,” she says.

A study published in January 2025 by Paige Nong at the University of Minnesota found that around 65% of US hospitals were already using AI-assisted predictive tools. Of those, only two-thirds evaluated their accuracy. Even fewer checked for bias. The number of hospitals using these tools has likely grown since then. But the evaluation gap remains.

Wiens isn’t calling for a halt. She believes in AI’s potential to improve clinical care. What she wants is more information — real, rigorous assessment of how these tools affect people in real clinical settings. “I have to believe that in the future it’s not all AI or no AI,” she says. “It’s somewhere in between.”

That sounds about right. But right now, we’re flying blind.