Clinicians embrace AI tools, but hospitals lag on training, governance

A global survey of nearly 2,800 clinicians finds AI is now part of everyday healthcare, but weak trust, limited training and poor governance threaten its wider adoption in clinical practice

Artificial intelligence (AI) has moved beyond experimentation to become a routine part of clinical practice. Nearly half of clinicians worldwide now use AI tools at work and half say the technology is already improving patient care. However, clinicians remain cautious about relying on AI in healthcare, as trust, governance and training remain the biggest barriers to its wider adoption.

 

These are among the key findings of Elsevier's Clinician of the Future 2026 report, based on responses from 2,757 doctors and nurses across 118 countries between December 2025 and February 2026. The report suggests healthcare has entered a new phase where AI is no longer viewed as an emerging technology, but as a tool increasingly integrated into everyday clinical processes.

 

AI is already part of clinical work

The report found that 49 per cent of clinicians now use an AI tool for work purposes, while 50 per cent believe AI is already improving patient care.

 

At the same time, AI adoption appears to have stabilised after the rapid growth seen over the past few years. According to the report, healthcare is entering a new era in which AI is becoming a standard component of clinical practice rather than remaining just a technology. The findings also suggest clinicians are increasingly using AI to streamline workflows, access medical information more quickly and support decision-making in complex cases.

Trust remains the weak link

Despite wider adoption, confidence in AI remains limited. Only 37 per cent of clinicians said AI tools are trustworthy, making trust one of the weakest indicators measured in the survey.

 

The findings show a growing disconnect between usage and confidence. Clinicians are increasingly incorporating AI into their work, but many remain unconvinced that current systems are sufficiently reliable for high-stakes clinical settings, where decisions can directly affect patient outcomes.

 

Trust levels also vary across regions. India (43 per cent), China (49 per cent) and Japan (54 per cent) reported higher-than-average trust in AI tools, while clinicians in the United States (19 per cent) and the United Kingdom (21 per cent) were among the least likely to trust them.

Generalist AI dominates current use

The report shows that clinicians continue to rely more heavily on general-purpose AI platforms than specialised medical systems.

 

Among clinicians who use AI for work, 56 per cent said they use generalist AI tools, such as ChatGPT and Microsoft Copilot, frequently or always. These tools are primarily used “to search medical literature, identify drug information, support professional education and prepare patient education materials”.

 

While Doctors are most likely to use generalist AI for medical research, querying medical literature and identifying drug information, nurses mainly use it for upskilling and patient education.

Clinical-specific tools are gaining ground

Among clinicians using AI at work, 34 per cent now use clinical-specific AI tools frequently or always, up from 22 per cent a year earlier. These tools are increasingly supporting more advanced clinical tasks, including clinical decision-making, second opinions on complex cases, patient monitoring and patient-care decision support.

 

Clinicians, thus, rated clinical-specific AI systems more favourably than general-purpose tools across every major measure, including usefulness, reliability, authority, safety and whether the information provided is current.

 

The report argues this shift highlights the evolution of AI from a productivity tool towards a "critical assistant" capable of supporting evidence-based clinical decisions.

Institutions are not keeping up

Even as clinicians increasingly adopt AI, healthcare institutions appear to be lagging in providing the support needed for safe implementation.

 

Only 41 per cent of respondents rated their institutions positively for providing access to digital tools, 40 per cent gave positive ratings for AI governance and just 32 per cent said their organisations provide good AI training.

 

The report also found that clinicians continue to face mounting pressures more broadly. Nearly two in five said exhaustion had impaired their ability to treat patients effectively, while high patient volumes and administrative burden remained among the biggest contributors to workload.

Clinicians expect AI to assist, not replace them

Around 80 per cent of clinicians expect AI to become a critical assistant in decision-making or at the point of care over the next five to 10 years, while 79 per cent believe AI skills will become an essential component of medical education and clinical training.

 

However, clinicians also made clear that wider adoption depends on stronger safeguards. The most important factors influencing confidence were AI systems that are easy to use (64 per cent), comprehensive (63 per cent), safe (62 per cent), transparent (61 per cent) and built on high-quality evidence-based sources (60 per cent).