Willing but not yet ready, what new research tells us about AI in rehabilitation

A new international study of almost 2,500 rehabilitation professionals found that most are already using artificial intelligence (AI) and intend to keep using it, but feel only moderately prepared to do so safely and critically. The findings point to a clear need for practical AI education that goes beyond general awareness and helps clinicians evaluate AI outputs, protect patient data and keep care person-centred.

If you have used an AI tool to draft a patient information sheet or summarise a clinical guideline, you are in good company. In a study just published in Disability and Rehabilitation, researchers from Radford University and Physiopedia surveyed rehabilitation professionals from every major world region who use Physiopedia Plus (Plus) for online learning. Of the 2,356 people who answered the question, 1,720 (73%) had used AI for some purpose and around half had used it at work. Large language models such as ChatGPT were the most commonly used tools, followed by digital assistants and documentation tools.

The study, led by Shala Cunningham with Lucy Aird, Tarina Van der Stockt and Md Warasi, used two established measures. The Medical Artificial Intelligence Readiness Scale (MAIRS-MS) captures how prepared people feel across four areas: understanding AI concepts and data, applying AI in practice, anticipating its future role and handling the ethical issues it raises. The Unified Theory of Acceptance and Use of Technology (UTAUT) explores what shapes a person’s intention to use a technology, including how useful and easy they expect it to be, encouragement from others, available support, confidence and anxiety.

Intention to use AI was among the highest scoring areas of the survey, while overall readiness was moderate, with an average score of 74.8 out of 110. Respondents felt most confident with basic statistical concepts and the ethical use of health data. They felt least confident explaining how AI systems are trained and organising their workflows to accommodate AI. Encouragement from colleagues and organisations scored lowest of all, and many people reported having the knowledge and resources to use AI but not the support to troubleshoot when things go wrong. In short, willingness appears to be running ahead of preparation.

The team then looked at what was linked to stronger intention to use AI among 850 respondents with complete data. Perceiving AI as useful for your work was the strongest factor, although its effect levelled off at higher scores. A positive attitude, greater readiness, confidence in your own ability, previous AI use and older age were also linked to stronger intention, while anxiety about AI was linked to weaker intention. Ease of use, social influence and access to support were not independently linked once these other factors were considered. Compared with physiotherapists, physicians and community-based rehabilitation workers reported lower intention.

Respondents from middle-income countries scored higher than those from high-income countries on several measures, including perceived usefulness and attitude towards AI. One possible explanation is that clinicians working with workforce shortages and limited services see AI as a way to extend access to rehabilitation. However, these comparisons were descriptive and the sample was digitally engaged, so they should not be read as evidence that some settings are better prepared than others.

The authors argue that readiness is not just about being comfortable with technology but about being able to judge whether AI outputs are reliable, free from bias and appropriate for the person in front of you. In practice, that might mean pausing the next time you use an AI tool to ask:

1. whether the output could reflect bias against particular groups such as those with disabilities
2. whether it fits this individual person’s goals, context and functional needs
3. what patient data are you sharing

The study also makes the case that individual learning is not enough on its own. If you lead a team or service, clear guidance on appropriate use, privacy, documentation and accountability will help translate your colleagues’ interest into safe practice. For educators, the message is to move beyond introductory awareness towards applied skills such as critical appraisal, recognising bias and integrating AI into clinical workflows.

As with any survey, the findings have limits. Respondents were online learners who may be more interested in technology than the wider workforce, all measures were self-reported and the cross-sectional design means the results show associations rather than cause and effect. The readiness scale was also originally developed for medical students rather than rehabilitation professionals. You can read the abstract of Readiness for artificial intelligence adoption among rehabilitation professionals in Disability and Rehabilitation.

Taken together, the findings suggest that rehabilitation professionals are open to AI but do not yet feel fully prepared to use it in practice. Closing that gap will take more than access to new tools. It calls for education that builds practical skills and confidence, alongside organisational support that makes safe use part of everyday practice. The authors also highlight important questions for future research, including whether readiness and intention translate into actual use and how AI affects clinical decision making, communication with patients and outcomes that matter to people with disabilities. Wherever you are on your own AI journey, the message is that your clinical reasoning remains central, and building your readiness now will help you use these tools in ways that benefit the people you work with.