GPs & Generative AI: Cautious Optimism at the Front Lines of Care

Host 1: You've probably heard about AI making waves across countless industries, right? Streamlining logistics, sparking creativity, but what happens when this uh powerful technology steps into somewhere really personal, like the doctor's consultation room? Today, we're doing a deep dive into, well, a critical question. How do frontline medical professionals, specifically UK general practitioners, you know, GPs, how do they really feel about generative AI? Our mission today is to unpack this big national online survey, we're talking over 1,000 UK GPs surveyed back in January 2025. And this isn't just about the tech itself, it's uh it's profoundly about the human element in medicine, the core of patient care and what it means for you listening in. We'll be digging into the key insights, their cautious optimism, but also frankly their persistent skepticism about these AI tools that are evolving so fast.

Host 2: Yeah, what's really fascinating here, I think, is that this survey gives us a proper snapshot of opinion at such a crucial time, just as these generative AI tools are, you know, really starting to filter into clinical settings. So it's vital we understand not just what these doctors think, but why their views matter so much for the future of the healthcare we all rely on.

Host 1: Okay, before we get too deep into the opinions, let's just quickly level set on what we mean by generative AI here. We're not talking about just any old computer program, are we?

Host 2: No, not at all. We mean things like large language models, LLMs. So, think of the general chatbots maybe you've used, like ChatGPT. But the survey also touches on more specialized tools coming into healthcare. Things like, um, ambient AI scribes, Nuance DAX is an example. They can basically listen to a consultation and write up the notes. Or clinical chatbots like OpenEvidence, which are trained on huge amounts of medical data. Largely, these tools create new content, text, summaries, patient info based on the vast datasets they've learned from.

Host 1: Right, creating new content. Got it.

Host 2: Precisely. And when you look at the survey results, the appeal for GPs is, well, it's pretty clear. They see real potential for AI to help with some of their biggest headaches. For instance, a really substantial number, 69% of GPs agreed AI could improve documentation.

Host 1: 69%. Wow. That's high. Why documentation specifically? What makes that such a target?

Host 2: Well, it's kind of notorious, isn't it? Documentation is a massive time sink for doctors. It often eats up hours they could be spending, you know, face-to-face with patients. So, the idea that AI could streamline that, maybe auto-transcribing, summarizing consultations, that's a really powerful draw. And it wasn't just documentation. 57% also saw potential for improving how patient information is gathered. And nearly 60%, uh 59%, anticipated increased efficiencies in healthcare overall.

Host 1: Okay, so efficiency, less paperwork.

Host 2: Exactly. And if you link that to the bigger picture, reducing that admin burden, it's known to be critical in fighting clinician burnout. So these perceived practical benefits for tasks like notes and data collection, they offer this sort of glimmer of hope for a less admin-heavy job, and that's a strong motivator, naturally.

Host 1: So it sounds like the initial promise is really about freeing up doctors from the, let's say, the more tedious parts of the job, letting them focus more on the patient. But, uh, that promise often bumps up against reality, doesn't it? Where did the survey show those sticking points, the skepticism?

Host 2: Yes, that's exactly where the caution really comes in. While many doctors see the potential for efficiency, they also voice some pretty profound concerns about AI's limits, particularly when it comes to the, well, the core human aspects of medicine.

Host 1: Right, the human element. And empathy specifically seems to be a huge sticking point. A really striking 71% of GPs in this survey disagreed that AI could enhance empathic communication. 71%! And that's not just a strong feeling, it's actually more skepticism than in a similar survey in 2024 where it was 65%. They seem pretty clear: empathy isn't something AI can deliver, in their view.

Host 2: Hmm, and this really flags a fundamental question, doesn't it, about the limits of technology in these deeply human interactions. Being a GP isn't just diagnosing or documenting, it's fundamentally about connecting, understanding, providing comfort. So when you see a majority of doctors expressing such strong doubts about AI fostering empathy, it signals a real concern about preserving the very essence of that doctor-patient relationship.

Host 1: And it's not just about whether the AI itself can be empathetic. GPs also seem worried about how patients might use it. What came out there?

Host 2: Yeah, that's another big one. A significant chunk, 59%, expected that patients might start relying on AI tools instead of actually going to see a doctor, which is a legitimate worry, isn't it? Could have serious health consequences if people delay getting proper medical attention. And adding to that, a majority, 60%, also disagreed that AI would increase patient privacy. That points to some deep-seated worries about data security. You know, who gets access to this incredibly sensitive patient data? It taps into that broader unease we all feel about how our personal info is handled by big tech systems.

Host 1: Okay, now here's something I found really interesting, a bit of a paradox. While loads of GPs expected AI to boost efficiency, the reality for those who'd actually used generative AI, well, it was pretty different. Only 17% of the GPs who'd used these tools said it actually reduced their workload. 17%! That's a massive gap between hope and experience. What's going on there?

Host 2: It tells us that the journey from potential to uh practical benefit is really complex. It's not always a straight line. While AI might speed up certain specific tasks, that doesn't automatically mean less overall work in a busy clinical setting. You know, in some cases it might even create new pressures, maybe an expectation to cram in more patients, or maybe they have to spend time checking the AI's output, which kind of cancels out the time saved. That discrepancy really highlights how carefully we need to think about implementation, about the whole workflow impact.

Host 1: Mhm. And then there were the areas where GPs just seemed unsure. Quite a high number of don't know answers for some critical questions, wasn't there? Like AI's impact on care inequities, 32% said don't know; the risk of errors, another 32%; potential to reduce patient harm, 33%. What does that big block of don't know tell us?

Host 2: It's really telling, isn't it? It reflects a significant level of uncertainty and maybe just a lack of deep understanding or hands-on experience for many GPs at this stage. Where people did have an opinion, those views were often quite polarized, which suggests that for a big segment, the jury is simply still out on these really profound implications. This uncertainty just underlines how urgently we need more clarity, more evidence as these technologies mature and uh get woven into practice.

Host 1: Yeah, it's clear the opinions are really diverse, sometimes conflicting, which brings us neatly to the human factor again, doesn't it? Especially around training and support. What did the survey find about the demand for training?

Host 2: Oh, it was overwhelming. Absolutely highlighted an overwhelming need. A huge 79% of GPs agreed they'd need more support or training to properly understand these generative AI tools. And that number is consistently high. It mirrors 80% from the 2024 survey. So this isn't some new or passing fad. They are actively asking for help. They recognize there's a big skills gap here. NHS Health Education England, you know, the part of the NHS focused on workforce training, they've identified this too.

Host 1: Okay, but here's the bit that seems really concerning. Despite that huge demand, only 11% of GPs said their employer encouraged them to use AI tools. 11%! And even fewer, just 5%, had actually received any formal training. Doesn't that point towards a sort of undercover use where GPs are maybe trying things out informally on their own without clear guidance or oversight? What are the risks there?

Host 2: Well, that lack of sort of institutional governance, it creates huge uncertainty around trust, around risk, around how you actually integrate these powerful tools properly. If you don't have clear policies or liability frameworks or just that organizational backing, doctors might lack the structural support they need to feel confident really adopting these things. This lack of formal support, it could seriously hinder safe and effective integration, and it makes it much harder to figure out best practices for using AI responsibly.

Host 1: So the formal channels seem lacking, but it sounds like even a little bit of exposure, maybe informal learning or minimal training, actually made a difference to confidence levels. What did the survey show about the GPs who had received some training, however limited?

Host 2: Mhm, yes, it absolutely did. The survey found that GPs who had received some form of training, they showed um noticeably greater confidence in AI's ability to enhance patient information gathering. That was 83% for the trained group versus 72% for the untrained. They also had more confidence in AI-enhanced patient communication, 69% versus 53%. And interestingly, this confidence extended to their own perceived need for more training. The trained GPs were actually less likely to agree that they needed more support and training. 84% felt they needed more, compared to 91% of their untrained colleagues. So it really suggests that even a small amount of structured guidance can significantly boost a GP's confidence and maybe make the technology feel less overwhelming.

Host 1: Right. So it's not just a simple AI, good or bad, question. It's heavily influenced by how different people perceive it and, crucially, how much support they get. Now, beyond training, the survey also uncovered some really fascinating differences in opinion based on things like gender and age. What can we draw from those demographic splits?

Host 2: Yeah, these differences are more than just interesting stats, I think. They hint at maybe deeper perceptions and priorities. For example, male GPs, generally speaking, were more likely to see AI as beneficial across a range of different tasks. So, 59% of men thought AI would improve diagnostic accuracy, compared to 41% of women. You see similar gaps for patient communication, 59% of men versus 48% of women saw benefits. Even for things like conveying empathy, 17% versus 12%; personalized treatment plans, 65% versus 57%; prognostic accuracy, 64% versus 58%. And more male GPs also thought AI tools would decrease patient harm, 51% versus 39%, and increase patient privacy, 17% versus 10%.

Host 1: Wow, that's a pretty consistent pattern across quite a few areas where men saw more potential benefit. What about the women GPs, then?

Host 2: Well, conversely, female GPs more often stressed the need for training. You had 93% of women agreeing they needed more support or training, compared to 89% of men. It suggests the perception of AI isn't uniform, you know. It's shaped by maybe professional roles, maybe even societal expectations. Perhaps women GPs, who maybe often prioritize the relational side of medicine more, are just more cautious about AI's impact on things like empathy and privacy. It implies that any strategies for engagement and training need to be tailored, address the specific concerns of different groups, not just a one-size-fits-all approach.

Host 1: That's a really important point about tailoring strategies. Now, age also seemed to matter in how GPs viewed AI, didn't it?

Host 2: Absolutely. The survey showed older GPs, those 56 or older, were actually more likely to think AI would improve diagnostic accuracy. That was 71% of them agreeing, compared to just 58% for the younger GPs, those 35 or under. And they also saw more potential for improved prognostic accuracy: 69% of the older group agreed versus 56% of the younger ones.

Host 1: Hmm, interesting. What might be behind that? Older, more experienced GPs having more confidence in AI for those complex judgment calls?

Host 2: It's possible. Maybe older GPs, with all that experience, see AI more as a powerful assistant. Something that can augment their own diagnostic process, handle routine data analysis maybe, allowing them to focus their expertise on the really tricky cases. But then, it was the younger GPs, 35 or under, who showed more concern about AI potentially increasing inequities in care delivery. 61% of them agreed with that, compared to only 44% of the over 56s. That suggests maybe a generational difference in perceiving the broader societal impacts, the ethical side of AI. Perhaps younger professionals are more attuned to how technology can worsen existing disparities if we're not careful.

Host 1: So we've got these varying views, this huge demand for training and this frankly striking lack of formal employer support. It paints a really complex, dynamic picture, doesn't it?

Host 2: It really does. It just highlights this critical need for structured education, for robust support systems. We need to equip clinicians with the skills and the confidence to use AI effectively and address those underlying concerns about equity and trust. Because without that support, the kind of informal adoption we seem to be seeing, well, it could lead to inconsistent use, maybe even make existing disparities worse. It really underscores the need for proper frameworks, ones that don't just push out information, but actively help clinicians think critically and integrate these tools responsibly.

Host 1: And this whole landscape is just evolving incredibly fast. It's not just doctors grappling with AI, is it? Patients are right there, too, often ahead of the game, even. We see studies showing patients are increasingly turning to AI for health advice, like uh one in 10 Australians reporting asking ChatGPT health questions.

Host 2: That's a really significant trend, and you can see the appeal, can't you? AI chatbots might offer faster advice, maybe tailored, often easier to read than trawling through complex medical websites. So this parallel patient adoption, it inevitably shapes how GPs themselves perceive and interact with AI, because they're encountering patients who are already using these tools, bringing that information into the consultation.

Host 1: And the tools themselves are getting more sophisticated all the time. Within the NHS itself, we're seeing these more advanced, medical-grade AI tools emerging, things like Heidi, Heidian Wright, plus those ambient scribe technologies we mentioned, and clinical chatbots like OpenEvidence are becoming more common. These developments are probably shifting GP perceptions over time, too, as they encounter AI that's more robust, more purpose-built, maybe addressing some of those earlier worries.

Host 2: Hmm, and this constant evolution just underlines why we need continuous evaluation, clear ethical frameworks. It's so crucial. The survey itself pointed to several really critical areas for future research, didn't it? Like, what specific kind of training do clinicians actually need to feel confident and competent? How do GPs see their professional responsibilities changing as AI becomes more embedded? What are patients' wider expectations? What do they want from AI in their healthcare? And, critically, how does a GP's own technical understanding, or lack of it, influence how they view and adopt these tools? These are the big questions we need to keep asking.

Host 1: So bringing all this together, what's the core takeaway about how different people see AI and just how vital that support structure is for adoption, and, ultimately, what does this tangled picture mean for us, the patient, and the future of our healthcare?

Host 2: I think the main thing is that integrating AI isn't like flipping a switch. It's not a single event. It's an ongoing evolution. And it absolutely requires constant checking, clear ethical rules, and really strong support systems. It's about moving from the state of, you know, cautious optimism mixed with real skepticism towards something more informed, more confident, and, ultimately, an integration that's truly beneficial for both the clinicians and the patients. The future of care looks like a partnership, doesn't it? Between humans and technology, yes, but also between patients and their evolving medical teams.

Host 1: Okay, so recapping the key points: We've seen UK GPs are, let's say, cautiously optimistic about generative AI for specific things like documentation, information gathering. And their belief that AI can improve efficiency in documentation has actually gone up since 2024. But there's still deep skepticism hanging around, especially about empathy, about patient privacy, and about whether it actually reduces workload in the real world. And that overwhelming demand for training stands in really stark contrast to the lack of formal employer support, which seems to be creating this environment of, sort of, informal AI use and quite a bit of uncertainty. So, given how fast AI is evolving and how clearly doctors want training and guidance, it leaves us with a thought, doesn't it? How do you, listening now, envision that ideal partnership between human clinicians and AI in your own healthcare journey, and what steps do you think are the most crucial ones to take right now to make sure this partnership is safe, ethical, and genuinely beneficial for everyone involved? Something to perhaps ponder until our next deep dive.