Once in a while, a scientific paper poses a question that we already know the answer to. That’s the case with a new study in Nature Medicine titled, “Is AI actually improving healthcare?”

Yes, it is. This much we know. The study also offers a big caveat: “In many cases, we do not know.” So many AI tools are so new that it’s still unclear whether they improve patient outcomes.

This distinction is essential. In health care, we have a tendency to treat AI as a monolith — a single, mysterious force. But asking “Is AI actually improving health care?” is a bit like asking “Do lasers improve surgery?” In the hands of a skilled surgeon using a validated tool, they allow for lifesaving precision; in an unproven setting, the question of efficacy is still open.

As the health care industry undergoes a “medical AI revolution,” with three-quarters of U.S. health systems using at least one AI application, it’s crucial to avoid speaking in generalities. Unfortunately, this happens often and creates a fog of skepticism. MIT Technology Review, in writing about the new study, had this headline: “Health-care AI is here. We don’t know if it actually helps patients.”

A growing body of concrete evidence demonstrates that some validated AI tools are already improving health care outcomes. For instance, studies have shown that AI-powered monitoring systems are currently helping nurses detect subtle physiological changes — such as early fever or pain indicators — well before traditional methods. These aren’t just incremental gains; they lead to faster interventions that shorten hospital stays and prevent complications.

We see similar success of AI-powered systems in the real-time detection of sepsis and acute kidney injury, as well as improved blood flow rates in stroke patients. A study found that implementing an AI tool that detects blood vessel blockages reduced the critical time interval — how long it takes patients to get treatment once they arrive at a hospital — for stroke patients by 86.7 minutes on average. Those patients also saw significant improvements in reperfusion rates.

As the Cleveland Clinic’s chief AI officer Ben Shahshahani says, “AI is no longer an experiment. It’s a real, scalable tool that can support patients, providers, and health systems — improving outcomes, reducing stress for caregivers, and making care more accessible and efficient for everyone involved.”

For example, the clinic uses AI programs that analyze images the moment they’re taken, saving time in the triage process. It also uses AI tools to “analyze massive amounts of patient data — from brain scans to genetic profiles — and develop more personalized care plans based on what’s worked for similar patients before,” the clinic says.

AI can have a wide array of positive impacts. But those impacts are not automatic. Across the health care landscape, we should all be wary of “black box” claims and demand rigorous evidence before assuming that a new piece of technology will pay off. AI should be treated carefully. Its findings can be wrong. It can miss important signs. 

In my work scaling health tech globally, I see and appreciate the scientific rigor required to create tools that can help save lives. Before the U.S. Food and Drug Administration — or authorities in other countries — approve a new piece of equipment, a great deal of testing and diagnostic work must take place.

When people hear claims like “we don’t know” whether AI helps patients, they can get the impression that all health care AI is unproven. Given how much confusion there is about this new field, these kinds of one-size-fits-all statements can slow down adoption of tools that work.

Health care has a long history of meeting breakthrough technology with a healthy dose of skepticism. When X-rays first became available, they were scary. But unlike AI, X-rays included physical dangers from radiation. Creators of CT scans also had to overcome skepticism.

In some ways, AI is the latest technology to face this phenomenon. Be wary. Be skeptical. But doing so doesn’t require ignoring proven ways AI is helping.

I’ve found that it helps to demystify what we’re talking about. So I explain to people that health care AI is a broad term for an array of technologies that can use machine learning to improve various aspects of the health care experience. Clinical AI is a subset of health care AI, focusing on patient care. It involves clinically validated datasets and medical-grade diagnostics.

The field is also changing rapidly. We don’t know what future technologies will achieve. Longitudinal data is essential to assess long-term improvements in health care outcomes. Those studies are already underway.

The exciting promise of AI isn’t just better data — it’s also equity. The feedback my team gets from patients makes clear that their lives and health are already being improved by AI.

Some people live far from medical facilities that collect health care data, so they’ve been unable to get electrocardiograms with any kind of frequency. Now, they can get crucial diagnostic data at home using AI-powered hand-held devices.

Others can now get hospital-grade data from sites closer to home, because smaller clinics are using AI-powered equipment that’s much less expensive than the large devices big hospitals use. It’s a revolutionary shift.

Meanwhile, in hospital settings and elsewhere, AI-powered tools are now discovering minute changes in people’s health that could signal problems down the road. For example, AI can spot changes in electrocardiograms that are imperceptible to the human eye. As a result, patients and health care providers take action sooner to prevent cardiovascular disease. These are very real, practical, and lifesaving results.

It’s also incumbent on those of us in the health tech space to remain vigilant in seeking ways to refine these tools. The goals are always to save people’s lives and improve their health; everything else is secondary.

Fortunately, entrepreneurs are increasingly joining this field, and the global market for AI in health care is undergoing explosive growth. It was estimated at nearly $37 billion last year, and is expected to reach a half-trillion dollars by 2033.

The potential is exciting, and a bright spot for humanity’s future. We can feel optimistic without giving up the need for proof.

So while we don’t yet know all the ways AI might improve health care, we know that it can. In many ways, it already has.

Priya Abani is CEO of AliveCor, a health tech company that builds AI-powered cardiology tools for consumers, patients, providers, and payers.