Artificial intelligence in occupational therapy: Democratising expertise or deepening divide?

Introduction

At the 2026 World Federation of Occupational Therapy Congress in Bangkok, artificial intelligence (AI) was not a peripheral topic; it was at the heart of the profession’s conversation about its own future. That conversation is overdue. AI is already reshaping how occupational therapists work, reason, and communicate, yet the profession lacks the cohesive frameworks and shared standards needed to engage with it critically and confidently (Jozkowski, 2025). This editorial argues that occupational therapists cannot afford to remain passive in that conversation. We have both the values and the expertise to shape how AI is used and to ask the harder question: for whose benefit?

The knowledge problem

A recurring concern at the WFOT Congress was the relatively short average career span of practising occupational therapists and the implications of that for a profession dependent on accumulated expertise. Stover and Jacobs (2025) make the point directly: as the gap between rising demand for occupational therapy services and the availability of experienced practitioners continues to grow, the profession risks losing the depth of specialist knowledge that takes years to build. In the context of housing adaptations, where occupational therapists must hold in mind a complex web of clinical, legislative, ethical, and practical considerations simultaneously, that loss is not abstract; it has direct consequences for the people we serve.

At Foundations, the national body supporting Disabled Facilities Grant delivery in England, we have been exploring this challenge through our OT Wise initiative. We developed a structured reasoning tool designed to support occupational therapists working within the Disabled Facilities Grant legislative framework. The tool is not intended to replace professional judgement. It is designed to make the mental load of complex decision-making more manageable, prompting practitioners to work systematically through the relevant competing factors. It supports consistency of decision-making and helps guard against the unconscious bias that can enter professional reasoning when time is short and caseloads are high. It provides, in effect, a safety net. The OT remains in the driving seat. The AI holds the map.

Democratising expertise

The structured reasoning and accumulated expertise that currently sit with experienced practitioners could, if approached thoughtfully, be made more widely accessible to less experienced practitioners, to people designing their own environments, and ultimately to disabled people themselves. But that possibility comes with an important caveat. AI is trained on the evidence and information available to it at a point in time, and the way it draws on and combines that information is neither neutral nor perfect. Democratising expertise through AI only works if practitioners and the people drawing on its outputs understand both its potential and its limits, and apply the kind of careful human judgement that guards against acting on information that is inaccurate or misleading. A recent scoping review confirms that AI technologies are already demonstrating improved access to services for people in remote areas (Bulan et al., 2025). That is a direction consistent with the profession’s core values, but one that requires us to travel it thoughtfully.

This vision is not without its own tensions. Digital access is not universal, and the risk of reproducing existing inequalities in a new form is real. Nevertheless, access to the technology itself is only part of the picture. As AI becomes more widespread, the more pressing question may be whether people have the knowledge and confidence to use it well and to recognise where it falls short. Democratising expertise through AI only works if people can engage with it critically, not just access it.

Navigating the hazards

The benefits described above do not come without serious risks that the profession must engage with honestly. Hallucination, where AI generates plausible-sounding but factually incorrect information, is a well-documented and significant problem. Alkaissi and McFarlane (2023) highlighted instances where AI produced fabricated citations and non-existent research, underlining the risk when clinicians accept AI outputs without independent scrutiny. In a high-stakes adaptations context, AI that confidently produces incorrect clinical or legislative information could cause real harm. Bias embedded in training data is an equally serious concern. Panch et al. (2019) define algorithmic bias as the application of an algorithm that compounds existing inequities in socioeconomic status, race, ethnicity, gender, disability, or sexual orientation, and amplifies inequalities in health systems. AI systems learn from the world as it is, not as it should be, and without careful design and ongoing scrutiny they risk reinforcing the very inequalities the profession seeks to address. These are not reasons to avoid AI, but they are strong reasons to approach it as a thought partner rather than an authority and to insist on transparency about how tools are built, trained, and governed.

A personal reflection

I want to add something more personal, because I believe it speaks directly to what is at stake for occupational therapists in how we engage with this technology. A core part of our role in adaptations is translating a detailed understanding of a person’s functional needs into a written report that can be understood and acted upon by housing colleagues who do not share our clinical frame of reference. That requires not only sound professional reasoning, but the ability to capture complex, nuanced, human experience on paper in a way that is clear, compelling, and accessible to a different audience. For me, that has always been the hardest part of the job. An informal screening assessment a couple of years ago indicated moderate dyslexia, something I had suspected since childhood but never had named. The fear of not being able to do justice on paper to what I understood about a person’s needs, of the gap between what I knew and what I could write, was a real and persistent source of anxiety throughout my career.

Working with AI as a writing partner has changed that. It has given me a way to bridge the gap between the clarity of my clinical thinking and the written word, so that the person’s story, their needs, their context, reaches the people who need to understand it. I am clear that the professional reasoning is mine. The knowledge of the person is mine. For me, AI has offered a way to work around a barrier that should never have been there in the first place, without compromising the professional reasoning that sits behind the words. I suspect there are occupational therapists, and perhaps the disabled people we work with, for whom AI offers a similar kind of liberation. That possibility deserves to be taken seriously.

A challenge for the profession

Occupational therapists are uniquely placed to contribute to the responsible development and implementation of AI in health, social care, and housing. Our expertise in activity analysis, our understanding of person-environment interaction, and our commitment to occupation-focused, person-centred practice are precisely the frameworks needed to evaluate these tools well. As Jozkowski (2025) argues, without clear competencies and shared frameworks the profession risks marginalisation and missed opportunities to uphold occupational justice. Action is needed across education, clinical practice, policy, and in the adoption of technology and its development. The profession needs to step forward in this conversation, not simply as end users of tools designed elsewhere, but as active shapers of how AI is designed, deployed, and governed. The question is not whether AI will change occupational therapy. It already is. The question is whether we will be in the room where those developments take place and where decisions are made.