Snapshot of UK physicians: Artificial intelligence in healthcare

Between 2-14 June 2025, the RCP carried out its second member snapshot survey of 2025. This included questions about AI in healthcare.

The majority (70%) of physicians were either very or somewhat supportive of AI tools being widely implemented in the NHS, but many are sceptical that the health system has the digital infrastructure required for it.

AI in clinical practice

• Most physicians are not regularly using AI in clinical practice: 16% said they use AI tools in their clinical practice every day, 15% weekly and 6% monthly. 33% said rarely and 30% said never. [571 respondents]
• AI is being used for a range of clinical tasks: when asked which clinical tasks were supported by an AI tool in their clinical organisation, the most commonly cited responses were radiology/pathology interpretation (42%), ambient AI for letters or notes in outpatient settings (29%) and AI to support clinical decision making (19%). The use of explicit LLMs within the EPR, and using AI to predict clinical changes like deterioration, did not attend (DNA) rates and discharge, were each reported by fewer than 23 respondents. [317 respondents]
• The NHS is not moving quickly enough to provide clinicians with NHS approved AI tools: almost seven in 10 (69%) said that they were using personal access to ChatGPT and Microsoft Copilot for clinical questions. 15% said that they were using a medical-specific AI tool for diagnosis and 21% were using a personal ambient AI tool. [305 respondents]Most physicians are not using AI to support non-clinical work: 64% said they did not use personal or trust AI systems to support non-clinical work, for example minute taking in meetings or AI drafting. 31% said they did and 5% said they did not know. [572 respondents]
• Physicians are optimistic about AI improving clinical care: 54% said that AI would improve the quality of clinical care. 17% said it would not change and 9% that it would worsen the quality. 20% said they did not know. [582 respondents]
• And optimistic about it improving their ability to care for patients: 54% said that AI would improve their ability to care for patients. 21% said it would not change their ability and 6% said it would worsen it. 19% said they did not know. [581 respondents]

Education and training

• We need to build physicians’ confidence in using AI tools: 31% said that they were somewhat confident in using clinical AI tools, 24% were not at all confident, 17% were neither confident nor unconfident, and 15% were somewhat unconfident. Only 8% of respondents said that they were very confident in using AI tools in their clinical roles. 5% did not know. [578 respondents]
• A majority said they need training in clinical AI tools: 79% said they needed training. 12% said they did not and 9% said they did not know. [551 respondents]
• The majority don’t have access to training in AI tools: When asked if they had access to training, 66% said no, and only 6% said yes. 28% said that they did not know. [548 respondents]

Benefits

• Physicians are supportive of AI in the NHS: 70% said they were either very (29%) or somewhat (41%) supportive of AI tools being widely implemented in the NHS. 10% were neither supportive or unsupportive, 7% were not at all supportive and 5% somewhat unsupportive. 9% said they didn’t know. [575 respondents]
• Physicians are hopeful AI will reduce administrative burdens and save time: When asked to select up to three biggest benefits of using AI in clinical practice, reduced admin burden (69%), time savings (62%), improved diagnostic accuracy (34%) were most commonly cited. 20% said better outcomes for patients, the same proportion that said cost savings. Improving health inequalities and there being no benefits accounted for 8% of responses each. [542 respondents]
• A majority of physicians believe AI will improve their working lives: 58% said AI would improve their working life. 15% said there would be no change as a result of AI and 9% said it would worsen it. 18% did not know. [577 respondents]

Barriers and concerns

• Physicians believe the NHS lacks the digital infrastructure required for AI: When asked if they believed that the NHS has the right digital infrastructure to support widespread introduction of AI that will make a difference, 48% strongly disagreed, 20% somewhat disagreed, 11% somewhat agreed and 3% strongly agreed. 8% neither agreed nor disagreed and 11% did not know. [548 respondents]
• When asked about main barriers to the deployment of clinical AI systems in the NHS, the most commonly cited response was lack of clinical expertise in AI (74% of 541 respondents), followed by the inability to integrate AI tools with other systems such as EPR (70%) and poor interoperability of systems (65%). Other barriers were a lack of trained IT professionals (55%), incomplete datasets (51%), lack of regulation (36%), too much regulation (18%). Other was selected 64 times. [541 respondents]
• Physicians are concerned about the risk of error: 73% said this was their biggest concern about using AI in clinical practice. Other concerns were:
  • liability risks (54%)
  • the risk of de-skilling clinicians (52%)
  • risk of model drift (meaning that the AI algorithm changes over time, 48%)
  • risk of bias (48%)
  • explainability risks (meaning that it’s not possible to know how the AI produces its output, 47%).
  • difficulty of consenting patients to use of AI (41%)
  • the risk of worsening health inequalities (24%)
  • that ‘my job will cease to exist’ (7%)

Other was selected 42 times (8%). [541 respondents]

The RCP has today published its new RCP view on digital and AI report , setting out recommendations for government, the NHS and others for the safe and ethical use of AI in healthcare.

If you are a physician and did not get a chance to respond to the survey, but would like to share your experiences with the RCP, please contact policy@rcp.ac.uk.