Best of 2025: The year's biggest AI-driven shifts in medicine

It's been a transformative year in healthcare, especially when it comes to AI. Healthcare AI spending reached $1.4 billion in 2025, according to recent analysis by venture capital firm Menlo Ventures. ^[]

Clinical documentation 

Stacey Lee, JD, associate professor of health policy and management at the Johns Hopkins Bloomberg School of Public Health, tells MDLinx that, first and foremost, healthcare workers continue to lean on AI to record, transcribe, and summarize appointments with physicians. Most meaningfully, these tools reduce the time doctors spend on documenting patient information. ^[]

Imaging interpretation 

Physicians are also using AI to quickly read and annotate imaging studies, says Lee.

A 2024 study found that AI can enhance accuracy and efficiency of interpreting medical images like MRIs, CT scans, and x-rays. ^[]

We’re seeing it in “areas of medicine where imaging, or visualization, is heavily used for diagnosis or patient prognosis,” she says.

Coding and billing automation 

AI systems have proven effective at reducing administrative tasks, according to Kang-Mieler. The hope is that AI will reduce coding and billing errors, ultimately improving billing practices and maximizing revenue for practicing physicians. ^[]

As Lee says, “AI is shaving hours off administrative work—work that rarely adds clinical value.” 

Patient engagement

That said, while AI has been shown to promote healthy behaviors and even prevent hospital readmissions, there are some challenges to overcome.

Some patients are hesitant to trust AI, research shows, and worry about the accuracy and data privacy. ^[] As a result, these concerns can delay patient adoption of AI. 

As for how healthcare workers feel about AI? “Most healthcare professionals fall somewhere between cautious optimism and controlled skepticism,” says Lee. Many have mixed feelings about using the tools during patient visits. 

“They see the relief—less charting, more time with patients—but worry about losing visibility into how AI reaches its conclusions,” Lee adds.

How HCPs feel about AI in 2025

As with anything, HCPs have differing opinions on how they feel about AI in medicine. Some, like Reddit user and primary care physician @Ivegotdietsoda, welcome the help AI can provide.

"I see it helping and improving my workflow, such as with notes (AI scribes), inbox, or other remedial tasks that should and can be automated. It can also very well replace sharing and explaining normal lab results, diagnostics, etc., and it should, I get bored with those things too," they wrote in a post on r/medicine.

Similarly, Reddit user and physician @doctorgreybc doesn't worry about AI replacing their job.

"AI is a tool, not a replacement," they wrote in r/medicine. "Yeah, it’s going to get really good at crunching data and spitting out differentials, but medicine isn’t just about knowledge. It’s about judgment, empathy, and navigating the messy, unpredictable parts of being human. AI can’t sit with a patient who’s scared, read between the lines of what they’re not saying, or make tough calls when the algorithm’s answer doesn’t fit the situation. That’s where we come in. The human element matters. Patients don’t just want a diagnosis—they want someone who listens, explains, and cares."

However, ER doc @nateisnotadoctor does fear their job will be replaced by AI in the future.

"[AI will] absolutely will replace me," they wrote in r/medicine. "I’m an ER doctor. I have a set of skills that in the current environment are probably about 80-90% replaceable by an appropriately trained AI, considering how metric driven and patient decentered we are pushed to be. I hate it, but pretending that we will never be replaced when the board certified emergency physician is ALREADY being replaced and supplemented by mid levels everywhere hospitals think they can get away with it is absolutely crazy."

Along the same lines, Reddit user @iplay4Him said: "[It's] hard to not see a world where 10 maybe 20 years from now every symptom we see (and every symptom some camera sees), is put into an algorithm that has millions-billions of patients' data backing it, and the differential and plan of action is just there. Sure I imagine we will have to approve it and actually do the thing at first, but it seems like it would make a lot of what we do and a lot of what makes all of this work worth it, unnecessary."