DR GPT™: Why I Built a Physician-Led AI Voice for the Future of Medicine
The story behind DR GPT™, told by the emergency physician who built it. Why healthcare AI needed a clinical voice, and what the brand stands for.
When ChatGPT arrived in late 2022, most of medicine reacted in one of two ways. A lot of physicians dismissed it as a toy. A lot of technologists announced that the diagnostic problem had been solved.
Both groups were wrong, and both were confident, and neither had been in an emergency department at 3 a.m. in a long time.
I wrote a book about it. ChatGPT and Healthcare came out in 2022, which made it one of the earliest attempts by a practicing physician to say something concrete about what these systems could and could not do inside real clinical work. Thirty-some books later, I am still writing about it, and the core argument has not changed much.
DR GPT™ came out of that argument.
The gap that started it
Two conversations were happening about healthcare AI, and they were not talking to each other.
In one room, engineers and founders discussed benchmarks, model performance, and market size. In the other, clinicians discussed workload, liability, and whether the thing in the chart was going to get somebody hurt. Both conversations were reasonable. Neither had a translator.
The decision to build DR GPT™ came from seeing that gap repeat in different rooms. Patients were already bringing AI-generated answers into care. Clinicians were already wondering what those answers meant for safety, trust, and liability. Leaders were already being asked to buy tools before they had a practical language for governing them. I wanted a physician-led voice that could speak across all three.
The name was almost a joke at first. Patients had already started saying they had asked "Dr. GPT" about their symptoms before coming in, which is either alarming or useful depending entirely on what happens next. I decided it should be useful.
What DR GPT™ actually stands for
Three commitments, and they have not moved.
A clinician in the loop, always. Not as a legal formality. As the person with the authority to say no, without a productivity penalty for saying it. Every framework I teach ends up back here, because every failure I have investigated traces back to a clinician who either was not consulted or was not free to disagree.
Evidence over enthusiasm. I cite journals on stage. That sounds obvious until you sit through a few healthcare AI keynotes. When I tell a hospital board that a widely deployed sepsis model showed an area under the curve of 0.63 and missed 67 percent of cases in external validation, I am pointing at Wong A, Otles E, Donnelly JP, et al., JAMA Internal Medicine, 2021, per PubMed (DOI). Named study, named journal, named year. The audience can check me.
Plain language, no theater. Healthcare AI is drowning in vocabulary designed to make simple things sound expensive. My test for any explanation is whether a night-shift nurse would recognize her own job in it.
Why a physician had to build it
Because the failure modes are clinical, and they are invisible from outside the building.
A model with excellent published performance can still be useless if its output arrives during a moment when nobody can act on it. A tool can be technically correct and still erode care by teaching a unit to ignore alarms. A dashboard can show green while the actual behavior on the floor has quietly changed. None of that shows up in a demo. All of it shows up on a shift.
I have also spent enough time on the other side of the table to know that clinicians are not automatically right about this. Physicians resist good tools for bad reasons all the time, usually because the tool threatens autonomy or adds three clicks. A credible voice has to be willing to say that too.
That is why DR GPT™ is not anti-technology and not pro-hype. The work is to ask whether a tool makes the clinical moment safer, clearer, faster, more humane, or more accountable. When it does, I want healthcare to move faster. When it does not, I want leaders to have the language and courage to stop.
Where the work sits now
The brand turned into a body of work rather than a product line, which was not the original plan and has turned out to be the right outcome.
Books, including AI in Emergency Medicine with Wiley and AI and Healthcare, 2nd Edition with Routledge. Keynotes and board sessions for health systems, societies, and government convenings. Advisory work, including Singapore's Ministry of Health Regulatory Advisory Panel and the Texas Medical Association's Committee on Health Information Technology, which represents more than 55,000 Texas physicians. Teaching, through the UTSA and UT Health Medicine and AI dual-degree program. And building, as Chief AI Officer at Phantom Space, where the problems look different and the tolerance for unvalidated systems is even lower than in medicine.
Different surfaces, one argument underneath.
What I am watching next
Three things.
Governance becoming real rather than rhetorical. The Joint Commission launched a voluntary Responsible Use of AI in Healthcare certification on June 2, 2026, covering governance, data management, risk and bias reduction, monitoring and validation, and transparency with education and training (Fierce Healthcare). Certification changes conversations that guidance alone never does.
The equity split in adoption. Ambient AI adoption ran at 70.2 percent among nonprofit hospitals against 28.8 percent at for-profit facilities, and 64.7 percent in metropolitan areas against 54.3 percent outside them (Yang F, Graetz I, American Journal of Managed Care, 2026). The hospitals with the least staff are getting the capacity tools last.
Spanish-language healthcare AI. I am a native Spanish speaker, and the gap between the volume of this conversation in English and its volume in Spanish is embarrassing given who receives care in this country and across Latin America.
In five years, I want DR GPT™ to be the physician-led reference point for practical healthcare AI. A place where boards, clinicians, founders, students, and patients can find plain-language guidance that respects both innovation and clinical reality. Not a slogan. A trusted voice for safer adoption.
The through-line
I did not build DR GPT™ because I think AI will save medicine. I built it because I think the decisions being made about AI in medicine right now are too important to leave to people who have never had to tell a family that something went wrong.
Physicians should be in that conversation. Loudly, with citations, and early enough to matter.
Harvey Castro, MD, MBA is a board-certified emergency physician, 5x TEDx speaker, and author of more than 30 books on AI and healthcare. He serves on Singapore's Ministry of Health Regulatory Advisory Panel, advises the Texas Medical Association's Committee on Health Information Technology, and is Chief AI Officer at Phantom Space.
Book Harvey for your board, leadership retreat, or healthcare AI event: www.HarveyCastroMD.com | /HarveyCastroMD | #DRGPT
Related DR GPT™ reading
- The Physician's Role in Responsible AI Adoption
- AI Governance in Healthcare: A Physician's Framework for Boards and Executives
More from DR GPT™: who DR GPT™ is, speaking, TEDx healthcare AI, books, media kit, and the healthcare AI board advisory.
Book Harvey Castro, MD, MBA, known as DR GPT™, for healthcare AI keynotes, board briefings, and leadership retreats. Check availability and book.