How are GPs leveraging AI? Insights from the front lines

Successive governments have sought to improve access to general practice, and thus the working lives of general practitioners, through a combination of technological, patient self-care, administrative, and staffing interventions. But newer is the idea that artificial intelligence (AI) will play a key role. AI refers to machines that perform tasks that typically require human intelligence, especially those that learn from data to find patterns and make predictions.

The research in this report aims to determine the proportion of UK GPs who are currently using AI in their clinical practice and explore how they are using AI. We sought to understand the scope of practice in which general practitioners use AI, the perceived benefits and concerns, and the barriers and enablers to wider adoption of AI in general practice.

AI refers to machines that perform tasks that typically require human intelligence, especially those that learn from data to find patterns and make predictions. This includes tasks such as transcribing audio of patient visits into summaries and patient notes, creating answers to patient questions, and tools to identify potentially serious skin conditions from images. We used a mixed methods approach combining a national survey (as part of the Royal College of General Practitioners' annual GP Voice Survey) and a series of online focus groups. The survey is the largest and latest survey of GPs on the subject and provides insight into the use of AI, just as the NHS 10 Year Care Plan promises to rapidly expand its use in the NHS.

Key findings from the survey include:

• Of the 2,108 respondents to the GP survey, 598 (28%) said they were currently using AI tools in their clinical practice. This figure breaks down to 13% of all GPs using tools provided by their practice, 11% using tools they obtained themselves, and 4% using a combination of both.
• There is wide variation in the adoption of AI among GPs from different demographic groups.
  • Male GPs were significantly more likely to use AI than female GPs. Of the 848 male practitioners who responded to the survey, a third (33%) said they use AI. This equates to a quarter (25%) of the 1,184 female GPs.
  • General practitioners working in socio-economically disadvantaged areas were less likely to use AI, especially clinic-provided AI. Just over a quarter (27%) of the 1,046 GPs who said they worked in more affluent areas said they used AI tools, compared to more than a third (35%) of the 467 GPs who said they worked in more affluent areas.
  • Additionally, GPs in England (31%) were more likely to use AI than GPs in Scotland and Northern Ireland (20% and 9% respectively). GPs in Wales were more likely to use AI (28%) than GPs in Northern Ireland (the difference with Scotland was not statistically significant).
  • Younger GPs were more likely to use self-acquired AI tools than older GPs. Of the 461 GPs under 35, 15% reported using AI tools brought into the workplace, compared to 11% of GPs aged 35-54 and 8% of GPs aged 55 and over. GPs aged 45-54 were more likely to use AI tools provided by their practice (15%) than GPs under 35 (11%) and GPs aged 65 and over (6%).
• Of the 597 respondents who reported how they use AI tools for tasks, more than half (57%) used AI tools for clinical documentation and note-taking. Around 4 in 10 GPs use AI tools for professional development (45%) and administrative tasks (44%), but fewer (28%) use AI tools to support clinical decision-making. However, focus groups suggested that GPs are actively testing AI tools for this purpose.
• General practitioners want to ensure that AI tools handle time-consuming, mundane tasks, allowing them to focus on complex clinical reasoning and meaningful patient relationships. For example, when asked what three AI development priorities they would like to focus on over the next two to three years, 2,108 GPs selected the following:

Regardless of their current use of AI, participating general practitioners at all career stages expressed the following concerns about the introduction of AI in general practice:

• Professional liability and medico-legal issues (89% of non-users, 80% of users using expert-selected AI tools, and 80% of users using home-grown AI tools)
• Lack of regulatory oversight regarding AI (88% among non-users, 78% among those using practitioner-selected AI tools, and 74% among those using home-grown AI tools)
• Risk of clinical error (83% for non-users, 69% for those using expert-selected AI tools, 70% for those using independently obtained AI tools)
• Patient privacy and data security (82% for non-users, 69% for those using expert-selected, independently obtained AI tools).

Focus groups revealed that the biggest benefits GPs experience from AI are time savings and reduced administrative burden. While policy makers would like this saved time to be used to offer more appointments, GPs reported using this time primarily for self-care and rest, including reducing overtime to prevent burnout.

Focus groups supported the findings. General practitioners highlighted the lack of regulatory oversight of AI as major concerns, as well as misleading or inaccurate output ('hallucinations'). Beyond ambient voice technology, for which guidance is being developed for the NHS at national level, the introduction of AI in general practice is likely to rely heavily on local policies developed by individual practices and integrated care boards (ICBs), and local staff willing to test tools and share learnings. However, the practice is inconsistent across countries, with focus group participants suggesting that some ICBs ban the use of AI altogether, while others actively encourage safe use and piloting. General practitioners emphasized the need for clear national standards, supported by training aligned with local policy.

To address variation in AI adoption and encourage responsible use, UK policymakers need to:

• Work towards early establishment Evidence-based national guidance Both administrative and clinical decision-making and generative AI tools should be covered to avoid the postcode lottery and address variations and inconsistencies between ICBs.
• immediately Clarifying professional responsibilities and safe use of AI, regulatory and governance frameworks. This should be done by a consortium of national policymakers, experts and sector regulators, and (similar to the guidance outlined above) should cover AI tools that are and are not considered medical devices.
• Comprehensive and structured development Training and education programs This is not just for those in medical education, but also for NHS staff in postgraduate studies. This should be nationally funded by the Department of Health and Social Care, and the regulator and the Royal College of Physicians should be involved in standardizing and specifying the content of this training.
• Using research on the impact of AI, Set realistic goals for the potential benefits of AI. While the 10-year healthcare plan suggests that AI will fundamentally improve patient access, this study highlights the need to recognize that some of the time saved will not translate into immediate increases in appointments, but rather reduced clinician overtime (and/or employee burnout).
• take action to Reduce the risk that the use of AI widens health disparities. This should also include addressing findings that AI users in poorer regions are less likely to have access to practice-based tools, and that some AI tools do not support minority languages.
• Consider environmental impact. The introduction of AI has the potential to increase carbon emissions and e-waste, contradicting the NHS and RCGP’s net zero targets. National guidance is needed to align the use of AI with environmental priorities.

These recommendations may also be relevant to policy makers in Scotland, Wales and Northern Ireland.

To ensure that future tools address the needs and concerns of GPs, AI developers and technology suppliers You will need:

• focus Developing AI tools to save GPs time It supports the automation of administrative tasks and clinical documentation in daily operations, and is not a substitute for clinical judgment.
• Integrate your tools seamlessly with your GP's electronic patient record Rather than a standalone or bolt-on tool.
• Coping with and alleviating hallucinations And we make sure that the training emphasizes that risk.
• Co-design tools for GPs Works with a diverse group of GPs and other practice staff.

Given that many GPs are already using AI, for the time being GP practicing with GP leader You will need:

• Develop interim local AI practice guidance until clearer national guidance, regulatory and governance frameworks are available. Encourage open discussion about all AI use within practices and consider developing clear local protocols covering patient consent, approved tools, and reporting of adverse events.
• Test, learn and share together. If possible, allocate time to evaluate tools, share learnings across the network, report problems with tools, especially those that led to errors in care, and share resources such as policies and case studies.
• Help educate patients about the practical use of AI. As AI becomes more integrated into general practice, it will be important to explain the role it plays in practice while maintaining the human element of care that patients value.

The government believes that improving access to and experience of general practice is a key priority and AI is expected to play a key role. Our research shows that AI has the potential to enhance patient care and reduce GP workload, but the benefits are not guaranteed and rapid adoption is not imminent. Currently, 28% of GPs across the UK (compared to 31% in the UK) use AI tools, but guidance varies widely. Some ICBs urge caution, while others encourage experimentation with approved tools. Concerns about regulation, liability and lack of national guidance remain major barriers.

Policy makers hope that AI will allow GPs to book more appointments, but they may need to reconsider their expectations, as most use the time saved to reduce overtime and burnout. Successful implementation requires addressing system-level issues such as clear guidance, training, equity, and protection of professional values. Policy makers and those involved in AI innovation need to act now.