Should We Be Segmenting Physicians Based Upon Their AI Usage?

A few weeks ago, I was interviewing a neurologist about his treatment approach in a rare disease. At several junctures during the conversation, he paused. Then I heard the faint tap-tap typing in the background.

Because he was off camera, I found myself wondering -- briefly, and perhaps unfairly -- whether he was answering an email... or had he typed my question into ChatGPT or Claude before answering? His responses to several questions sounded really polished, almost textbook in phrasing, and for a moment I was tempted to call attention to it. But then I stopped myself. Who knows? In a strange way, the more fascinating question was not whether he had used AI in that moment, but whether that moment reflected something that is increasingly a routine behavior in 2026 medical practice.

That is where my brain kept going after the interview. More and more physicians are now working with AI in some part of their workflow, whether they are conscious of it or not. Ambient dictation tools are becoming increasingly common. Clinical decision support is embedded in electronic health records. Evidence aggregators are becoming part of the background hum of busy medical practice. In some settings, a physician may use AI as casually as they once used a journal article, a calculator, or a quick curbside consult with a colleague.

The physician of 2026 may not always be “AI-first,” but many are clearly becoming AI-assisted.

‍

A New Kind of Clinical Habit

What strikes me as particularly interesting is that AI is beginning to reshape not simply what physicians know, but how they arrive at what they know. It is beginning to fundamentally re-shape how clinical habits are formed. If a physician routinely utilizes an evidence summarizer before arriving at a treatment decision; or relies on ambient listening to preserve time, attention and focus during the visit; or turns to an AI platform when faced with unknowns...those actions may gradually become part of the way that physician practices medicine. The AI tool is no longer just supporting the decision. It is becoming part of the decision-making architecture itself.

That has obvious implications for pharmaceutical marketing research. When we interview physicians today, we are not always talking to the same cognitive actor we were talking to 3-5 years ago. Some physicians are now practicing in a world where AI has already influenced their workflow, their speed of retrieval, their confidence in what counts as evidence, and perhaps even the structure of the questions they ask themselves. On the other hand, others are still practicing in a far more traditional way, relying primarily on memory, experience, journals, reps, and their own clinical judgment. Those are not trivial differences. They may represent two different behavioral species or archetypes.

‍

The Segments We May Be Missing

If that is true, then one of the most important things pharma can do is start to understand these physician segments more explicitly. There may be doctors who lean heavily into AI-supported decision-making, and others who are far more skeptical of it. There may be physicians who use OpenEvidence or similar tools routinely, and others who use them only occasionally, or not at all. There may be some who welcome ambient listening because it frees them to focus on the patient, and others who feel it subtly changes the texture of the encounter. These differences are probably not just technological. They may reflect deeper disparities in habit, openness to external inputs, tolerance for uncertainty, and comfort with delegation.

For commercial teams, that matters. A physician who is already accustomed to seeking machine-assisted input may respond differently to evidence, promotional messaging, and education than one who is still constructing most decisions from "first principles." The AI-assisted physician may be faster, more digitally fluent, and perhaps more comfortable with compressed information. The non-AI-assisted physician may be more skeptical, more self-contained, or more reliant on established routines. Neither group is any “better” than the other. They are simply using different frameworks for how evidence gets translated into action.

‍

What Pharma Should Be Studying Now

That is why I think pharma needs to study this phenomenon deliberately and systematically. We should be asking how physicians are using AI in their practices, how often they use it, for what kinds of decisions, and what effect it has on confidence, speed, habit formation, and willingness to change. We should probably be comparing physicians who use AI frequently with those who do not, not just to understand adoption patterns, but to understand whether AI is influencing the way they think about evidence and make decisions.

This is not a futuristic issue. It is already here. And like so many things in medicine, it is likely to matter most where habits are strongest and uncertainty is highest. If AI changes the rhythm of how physicians evaluate evidence, then it probably will also change how they receive promotional messages, how they process disease education... all the way up to how quickly they are willing to act on a newly approved treatment option.

That makes this a behavioral question as much as a technological one. And it is one I suspect we will be thinking about for a long time.