Exploring the Use of Generative AI in Postgraduate GP Training: A Qualitative Study of Resident Doctors’ Experiences and Perceptions

In the modern clinic, a doctor's day is a constant negotiation between the human need for connection and the machine-like demand for paperwork. Patients arrive with complex stories, but the doctor must also navigate electronic records, referral letters, and endless administrative checklists. Into this busy landscape has stepped a new kind of digital assistant: generative artificial intelligence. Unlike older computer programs that simply sort data or follow rigid rules, this newer technology can create original text, summarize information, and draft documents by learning from vast amounts of human writing. It is a tool that promises to handle the repetitive tasks that often crowd out patient care, yet it also raises a fundamental question: can a machine that writes and reasons be trusted in the high-stakes environment of medicine, especially when the person using it is still learning the ropes?

A team of researchers from Newcastle University and the North Tees and Hartlepool NHS Foundation Trust set out to answer this question by listening to the people on the front lines: general practice resident doctors. These are the doctors who have finished medical school but are still in their final years of training, working as fully qualified clinicians while simultaneously learning the specific skills of family medicine. They occupy a unique space, balancing the pressure of treating real patients with the need to study and reflect on their own growth. The researchers gathered nineteen of these doctors from the Northern region of the United Kingdom and held five group discussions to understand how they were actually using these new tools in their daily work, what they saw as the benefits, and what fears kept them up at night.

The conversations revealed that for these trainees, artificial intelligence had already become a practical partner for managing the sheer volume of paperwork. The most common use was to tackle the administrative heavy lifting. Doctors described using the technology to draft consultation notes, transcribe what was said during a patient visit, and write referral letters to specialists. One doctor noted that while they had their own way of documenting a visit, the tool helped enhance those notes, while another described a system that listened to the entire conversation and produced a full record, leaving the doctor to simply review the output. Beyond the clinic walls, the technology helped with the educational side of their training, organizing their professional portfolios and helping them map their daily experiences to the specific competencies required to qualify. It was not about replacing the doctor's work, but about speeding up the parts of the job that felt repetitive, freeing up time to focus on the patient.

However, the utility of the tool extended far beyond simple typing. The doctors also described using it as a thinking partner to help them navigate complex medical puzzles. When faced with a patient whose symptoms were unfamiliar, they would ask the system to generate a list of possible conditions or to explain a specific treatment pathway. This did not mean the machine was making the diagnosis; rather, it acted as a cognitive aid, offering a broad range of possibilities that the doctor could then filter through their own clinical judgment. It helped them prepare for exams, generate practice questions, and even simulate conversations to practice their communication skills. In this way, the technology served as a personalized tutor, helping them learn faster and understand difficult concepts by breaking them down into clear, conversational explanations.

Yet, this reliance on a digital assistant came with a heavy dose of caution. The doctors were acutely aware that the technology could make mistakes, sometimes confidently stating things that were simply wrong. They shared stories of the system misinterpreting a patient's story, such as confusing a human bereavement with the death of a pet, a type of error that could have serious consequences if not caught. This led to a firm consensus: the doctor must always be the one holding the pen. They emphasized that they had to verify every piece of information the machine produced, treating it as a draft that required expert review rather than a final answer. There were also deep concerns about privacy. The doctors worried that patient data entered into these systems might not be secure, potentially exposing sensitive medical histories to third parties. They stressed that using the tool without a patient's knowledge or consent felt ethically wrong, and that the responsibility for the care of the patient must always remain with the human doctor, never the algorithm.

Perhaps the most striking finding was the feeling of being unprepared for this new reality. While the doctors saw that artificial intelligence was going to become a standard part of their future careers, they felt that the training they received to handle it was inconsistent and often non-existent. They were learning by trial and error, trying to figure out which platforms were safe and which were not, without clear guidance from their institutions. They expressed a strong desire for formal, expert-led education that would teach them not just how to use the technology, but how to use it safely, how to spot its errors, and how to manage the changing dynamic when patients arrive with information they found from similar tools. The doctors recognized that the relationship between doctor and patient was shifting, with patients sometimes bringing fixed ideas based on what an algorithm told them, making the consultation more challenging.

The study concludes that while generative artificial intelligence offers a powerful way to reduce the administrative burden and support the learning of future doctors, it is not a substitute for human expertise. The technology works best as a supplement, a tool that handles the routine so the doctor can focus on the complex, but it requires a vigilant human hand to guide it. The researchers found that without clear, standardized training and professional guidelines, the risks of over-reliance, data breaches, and clinical errors remain too high. For the next generation of doctors to thrive in an AI-enabled future, they need more than just access to the tools; they need the wisdom to know when to trust them and when to look past them.