The Nuffield Trust finds that 28% of UK GPs use AI for practice tasks, but challenges such as regulation, safety, and inequality persist
A new report from the Nuffield Trust and Royal College of General Practitioners finds 28% of UK GPs already use AI tools in clinical work, mainly for documentation, admin, and professional development. AI saves time and eases workloads, but doctors remain concerned about regulation, patient safety, and uneven access.
AI use in general practice
AI can be used in GP practices for tasks such as transcribing speech from patient consultations into summaries and patient notes, producing responses to patient queries, and identifying potentially serious skin conditions in images.
Researchers combined a national survey and online focus groups to create the largest, most up-to-date survey of GPs’ use of AI.
Key survey findings
- Overall adoption: 28% of GPs (598 of 2,108) currently use AI tools in their clinical practice. Breakdown: 13% use practice-provided tools, 11% use self-obtained tools, and 4% use both.
- Demographic differences: 33% of male vs. 25% of female GPs use AI.
- Area deprivation: 27% of GPs in socioeconomically deprived areas use AI, compared with 35% in more affluent areas.
- Region: England 31%, Wales 28%, Scotland 20%, Northern Ireland 9%. Differences between Scotland and Wales were not statistically significant.
- Age: Younger GPs (<35) are more likely to use self-obtained AI (15%) than GPs aged 35–54 (11%) or 55+ (8%). GPs aged 45–54 are most likely to use practice-provided AI (15%), compared with those under 35 (11%) and 65+ (6%).
- Task usage: Among 597 respondents reporting on tasks:
- 57% use AI for clinical documentation and note-taking
- 45% use it for professional development
- 44% use it for administrative tasks
- 28% use AI for clinical decision-making.
GPs express concerns about AI’s adoption in healthcare
GPs, regardless of AI use, cite top issues as professional liability and medico-legal risk (89% non-users, 80% using practice AI, 80% self-obtained), lack of regulation (88%, 78%, 74%), clinical errors (83%, 69%, 70%), and concerns over privacy and data security (82% of non-users and 69% of AI users).
Focus groups say AI’s main benefit is saving time and reducing admin. GPs warn that the lack of regulation and risks of incorrect outputs are major concerns.
Currently, ambient voice technologies are the only tools with supportive NHS guidance, and the implementation of AI in general practice appears to depend heavily on local policies developed by individual practices and Integrated Care Boards (ICBs).
How can policymakers address variations in AI adoption?
Policymakers in England must act quickly to reduce variation in AI adoption and promote responsible use. They must set clear, evidence-based national guidance for administrative, clinical, and generative AI tools, and clarify liability, safe use, and regulation frameworks.
Structured training for medical students and NHS staff should be funded nationally, with regulators and the Royal Medical Colleges standardising content. Research should set realistic expectations: time saved may reduce clinician burnout rather than immediately increase appointments.
Action is also needed to prevent AI from widening health inequalities, including lower access in deprived areas and tools that don’t support minority languages. Finally, environmental impacts, such as carbon emissions and electronic waste, must be considered to ensure AI aligns with the NHS’s net-zero goals.
