AI adoption in UK general practice ‘unsupervised and fragmented’, finds study

Adoption of AI in general practice in the UK is happening in a ‘fragmented and unsupervised’ manner, researchers have warned.

A quarter of GPs in the UK now use generative artificial intelligence (GenAI) tools for clinical work, their data suggested. However, the researchers found that the vast majority of GPs have not received professional training on its use.

Their survey of more than 1,000 UK-based GPs found:

• 25% of respondents used GenAI tools in clinical practice, compared to 20% in 2024. 

• Out of all respondents, 95% reported they had no professional training in using GenAI tools at work. 

GenAI refers to models that are pre-trained on large datasets and use this to create original content based on a user’s request – for example, a GP asking it to summarise the transcript of a recorded appointment with a patient. 

Of the 249 GPs who said they used GenAI for clinical tasks, generating documentation after patient appointments was the most common use (35%, up from 29% last year). 

GPs also reported using GenAI for: suggesting a differential diagnosis (27%), treatment options (24%), and referrals (24%). 

In all, 71% of GPs who use AI tools said they reduced work burdens. 

Overall, 85% of respondents reported that their employer had not encouraged them to use GenAI tools, but only 3% said their employer had actively prevented them (‘outright prohibition’) from using GenAI tools in their work.  

In response to the findings, the study leaders from Uppsala University concluded that AI adoption in clinical settings in the UK ‘continues to occur in an unsupervised and fragmented manner’.

They recommended that ‘medical organisations and policymakers must develop guidelines, provide training, and establish clear regulatory frameworks.’

The Government made AI transcribing tools a key part of the 10-year NHS health plan, published earlier this year, committing to rolling out ‘validated AI diagnostic tools’ and deploy ‘AI administrative tools’ NHS-wide over the next two years. 

The plan also said that the NHS App will be expanded to include a new AI-enabled ‘My NHS GP’ tool to handle non-urgent care enquiries. 

Guidance from NHS England specifically on AI transcribing technology – or ambient voice technology (AVT) – suggests any GP staff using it should be trained. 

The document, intended for chief information officers (CIOs) and chief clinical information officers (CCIOs) to set policy, says they should ‘train staff on the appropriate and approved use’ of ambient scribing products.  

It suggests ‘providing guidance on appropriate dictation regarding voice and words used, and reinforcing ongoing responsibility for practitioners to review and revise the outputs’. 

NHS England has said transcribing software performing more sophisticated tasks like summarisation would likely be within the Medicines and Healthcare products Regulatory Agency’s (MHRA) remit as ‘medical devices’, with their enhanced role meaning staff using the technology are mandated to be trained with it. 

And MHRA has specifically warned GP practices of the risk of hallucination when AI tools are used for tasks like transcribing and summarising appointments. It explained there is a risk of ‘hallucination’ with AI products which users should be aware of, and asked GPs to report all adverse events or suspected inaccuracies to them. 

Last month, a World Health Organisation (WHO) report found AI’s increasing role in healthcare across Europe has not been accompanied by basic legal safety nets needed to protect patients and health workers. 

It found widespread use in helping doctors spot diseases, reduce administrative tasks and communicate with patients – but questioned systems for liability when an AI system makes a mistake or causes harm.