“I Felt Violated”: What One Patient Complaint Taught Me About AI in the Exam Room

After 25 years as an OB-GYN, I thought I had heard everything a patient could say to me. I was wrong. A phone call with a patient — prompted by a complaint about her recent visit — stopped me cold. What she shared about her experience with the AI technology used during her appointment was something I was wholly unprepared to hear. I have reflected on that conversation ever since.

The week before, a new patient came to me for a routine examination. Over the past year, my hospital has been rolling out an AI scribe. The app records the visit and then synthesizes notes based on the recording. Each patient must consent to its use, and up until this point, none of my patients raised any issues with the technology. However, this patient paused. In the end, she agreed and consented to the use of ambient AI. The exam went well, but several days later, my administrator told me someone had filed a complaint. I take complaints very seriously and decided to call the patient to find out what went wrong. That’s when she said it: “I felt violated and like the AI was in my vagina.”

Advertisement

As an OB-GYN, “I felt violated” are the three worst words we can hear. OB-GYNs spend years of their careers learning to be sensitive and to build patient trust. I had never received that kind of comment in my entire career — and now I had, because I used what I thought was benign technology.

While AI scribes are relatively new to the medical world, scribes have long been a part of patient care. Physicians have used medical students to take notes for longer than I have been practicing. During my career, I have also seen the implementation of virtual scribes — outsourced typists who join appointments remotely. Dictation services are common across healthcare practices. To me, AI scribes seemed like a natural progression of these services.

These AI scribes are incredibly helpful for practitioners. Not only do these apps handle dictation, but they also help synthesize notes and charts. Ask any physician, and they will tell you that documentation is a significant burden — often on our staff as well. This technology makes the often-overwhelming paperwork much easier to manage, allowing us to focus on patient care.

That may be why I never thought critically about using this type of AI until I was forced to. AI had come between my patient and me, and I had to take that seriously. I needed to learn more — especially since I serve on my hospital’s AI services rollout team. For the past year, I have worked closely with our AI services provider and hospital administration to bring AI into our institution. I help train other providers on its use and serve as a point person. Yet I had not been thinking critically about this technology.

This troubling interaction happened just before the holidays, when I spent two weeks with my three daughters, who include members of Gen Z and Gen Alpha. When I brought up the AI scribe, they were quick to raise concerns about the environmental effects of AI — specifically, the immense amount of water it consumes. According to scientists at the University of California, Riverside, each 100-word AI prompt is estimated to use roughly one bottle of water. I could only imagine how much fresh water would be consumed to transcribe a 30-minute examination. If that weren’t enough, AI data processing facilities are frequently located in or near low-income, predominantly Black communities, compounding environmental inequities.

These companies must know how much water they are using and where — they cannot afford for their data centers to overheat. Yet providers are not given this information. It does not surprise me that AI companies are not forthcoming about this when selling to physicians, but this is where my questions began to accumulate.

Practitioners must begin asking critical questions about AI usage in their practices. I have started to think carefully about AI, the environment, and who is most affected. Historically, we have rolled out new technologies without fully considering their repercussions. That approach does not work with sensitive medical technologies. For the time being, I have stopped using AI in my practice. It is not worth the risk of further eroding trust between my patients and me.

I want to be clear: I do not have the answers. I cannot offer a policy fix, a technical solution, or a roadmap for how medicine should integrate AI into the most intimate spaces of patient care. What I do have is a deep sense of professional responsibility. As a physician leader, I feel compelled to bring this issue to the forefront — to name what happened, to sit with the discomfort of it, and to invite my colleagues to do the same. Medicine has always advanced through the courage to ask difficult questions, even when the answers are not yet in sight. This is one of those questions.

While many patients simply say “sure” when asked to consent to AI use, they should also feel empowered to ask questions about the technology we employ. As companies roll out these tools, we must consider the patient, the provider, and the staff — and I do not believe that consideration has been adequate. It is easy to say that AI is “just taking notes,” but it is so much more than that. I hope that other providers do not need a patient complaint to recognize it.