Physicians increasingly recognise artificial intelligence as relevant to clinical practice, but routine use remains limited. An international cross-sectional survey published in npj Digital Medicine examined physician perspectives across 50 countries and territories, drawing on responses from 1,049 physicians collected between June and September 2024. The findings show broad awareness of AI and strong expectations that it can improve care, especially by making clinical work more efficient, timely and effective. Practical adoption is far less advanced. Many physicians have not used AI tools in healthcare, and most have not received formal AI training. The gap between interest and use appears closely linked to organisational conditions, including training, infrastructure, access to technology and institutional investment in AI-enabled tools.
Awareness Remains Ahead of Use
Physicians show a relatively high level of self-reported understanding of AI in healthcare. Most respondents describe their understanding as ranging from basic to advanced, and many also report at least some familiarity with AI applications. Confidence is more measured. Trust in AI for diagnostic and treatment decisions tends to cluster around moderate levels, rather than strong endorsement. Views are also divided on whether currently available AI products are well matched to clinical needs, suggesting that awareness does not automatically translate into confidence in available solutions.
Practical exposure remains limited. Fewer than three in ten physicians had used AI tools in healthcare, and formal training was uncommon. This creates a significant implementation gap: AI is widely recognised as relevant to clinical practice, yet many physicians have little direct experience with its use in real-world settings.
Ethical and privacy concerns are also prominent. Most respondents express at least some concern about patient data privacy in clinical AI. Among physicians with direct experience of AI tools, some had encountered privacy-related issues. These findings point to a cautious professional environment in which physicians see potential but remain alert to risks affecting patient data, clinical responsibility and trust.
Organisational Readiness Drives Adoption
The main obstacles to AI adoption are structural. Lack of infrastructure and resources is the most frequently cited barrier, followed by limited access to data and computing power, privacy and security concerns, ethical and regulatory issues, lack of trust and resistance to change. More practical difficulties also matter, including limited access to AI technologies and integration with existing systems. These barriers show that AI implementation depends on more than individual interest or professional openness.
Must Read: AI Trust Gap Tests Healthcare Adoption
Physicians who had already used AI tools also reported problems with performance in practice. Some said technologies failed to meet clinical expectations or created real-time difficulties. Adoption drivers reflect this experience. Demonstrated effectiveness and accuracy are the most frequently selected factors influencing uptake, followed by ease of use and integration into existing systems. Training, support, cost-effectiveness and regulatory approval also appear, but they are selected less often.
Formal AI training and institutional access stand out as the strongest adoption factors. Physicians with formal training were more likely to use AI, while those working in institutions that had purchased AI technologies were far more likely to report use. Other factors, including age, gender, workplace type, professional role and belief that AI tools are tailored to healthcare needs, did not show the same strength. This makes adoption primarily an institutional and capability issue, rather than a matter of enthusiasm alone.
Clinical Priorities Centre on Efficiency
Physicians connect AI with several persistent healthcare pressures. Limited resources and shortages of healthcare professionals are among the most common challenges identified, alongside poor infrastructure, limited access to services and difficulties in complex diagnosis and management. Within these pressures, AI is most often seen as a way to improve efficiency, timeliness and effectiveness. Patient-centred care and safety are also identified, while equity receives less emphasis.
Telemedicine and remote patient monitoring are the most widely supported AI-enabled responses to healthcare challenges. Decision support for diagnosis and treatment, disease surveillance and outbreak prediction, personalised medicine, resource allocation and electronic health record analysis also feature. Medical imaging analysis is viewed as the most useful AI-powered tool, followed by EHR management and patient monitoring. These priorities suggest that physicians see AI most clearly in areas where it supports clinical workflow, data handling, monitoring and decision-making.
Views on AI and inequality are mixed. Some physicians believe AI can reduce inequalities in healthcare access, while others are uncertain, expect inequalities to worsen or are unfamiliar with social determinants of health. Physicians also see healthcare providers as active contributors to AI development and implementation. Their roles include identifying clinical needs, taking part in design and testing, advocating ethical use, promoting awareness and collaborating with developers. Training, knowledge sharing, investment and clear frameworks are seen as important supports for integration.
Physicians across varied healthcare settings recognise AI as a relevant part of clinical practice, but routine adoption remains limited. The main gap lies between awareness and operational use. Training, institutional investment, infrastructure, access to deployed technologies and integration with existing systems shape adoption more strongly than general optimism or demographic factors. Effective implementation therefore depends on practical organisational support, clear attention to data privacy and ethics and meaningful involvement of physicians in the development, testing and use of AI tools.
Source: npj Digital Medicine
Image Credit: iStock
References:
Bold B, Serin O, Tantri Adhiatma L et al. (2026) Global physician perspectives on artificial intelligence in healthcare across 50 countries and territories. npj Digit Med: In Press.
