The study, by researchers at the Universities of Bristol, Cambridge, Oxford, Aberdeen and Queen Mary, University of London, explored what matters most to patients, members of the public and GPs when considering AI technologies designed to help detect skin cancer in primary care.
Researchers found that avoiding missed skin cancers was the most important factor for all groups, where a skin cancer is incorrectly identified as harmless.
The study also found strong support for developing and testing AI systems using data from people with a wide range of skin tones. Participants wanted to ensure that new technologies would work equally well across diverse populations.
More than 2,300 people from across the UK took part in the study, including GPs, patients and members of the public. Participants were presented with different AI skin cancer detection technologies and asked to choose between them based on characteristics such as accuracy, cost, where the technology would be used, whether it had been recommended in clinical guidelines and the range of skin tones it had been developed and tested on.
The researchers found that all these factors influenced people's choices, but accuracy was by far the most important consideration. The risk of missing a cancer mattered even more to GPs than it did to patients or members of the public.
The study also revealed that people preferred AI tools that are used by healthcare professionals in GP surgeries rather than technologies that patients use on their own at home, enabling patients to still speak to a human clinician about the skin lesion they are worried about rather than just get an autonomous AI result.
The findings come as interest grows in using AI in the NHS to support healthcare professionals and improve access for patients. Skin cancer is the commonest type of cancer, and AI technologies have been proposed to help detect skin cancer more quickly and efficiently.
However, the researchers say that successful implementation will depend not only on how well the technology performs, but also on whether healthcare professionals and the public trust it and the way the technology is implemented is acceptable to them.
Dr Owain Jones, GP and Academic Clinical Lecturer in General Practice at the Centre for Academic Primary Care, University of Bristol and lead author of the study, said:
"AI has the potential to support earlier detection of skin cancer, but our findings show that accuracy must come first. People want to know that these technologies will not miss cancers and that they work well for everyone, regardless of skin tone.”
What this means
The findings suggest that developers, healthcare organisations and policymakers should prioritise reducing missed skin cancers when designing and implementing AI tools for skin cancer detection. Equally, AI technologies should not increase workload for healthcare services.
The results highlight the importance of ensuring AI technologies are trained and tested using data that represent people with a wide range of skin tones. This could help build confidence that these tools work fairly and effectively for everyone. Participants preferred AI technologies that are in the hands of clinicians, ensuring that patients still have contact with human clinicians as part of their assessment.
As AI becomes increasingly integrated into healthcare, understanding public and professional expectations will be crucial to ensuring these technologies are accepted and used successfully.
Paper: Preferences for the use of Artificial Intelligence (AI) technologies to help detect skin cancer in primary care settings: a UK-wide discrete choice experiment (DCE), by Owain Jones et al. published in the British Journal of Cancer. September 2026.