will AI replace occupational therapists?

Of the 38 tasks analysed in your role, 37 show zero AI penetration. That's not a rounding error. It means the core of what you do every day — evaluating a stroke patient's grip strength, watching how a child with autism navigates a sensory activity, deciding whether a 78-year-old can safely return home after a hip replacement — is work that AI simply has no path into right now.

The judgment calls are yours alone. You're reading how a patient holds a fork, whether they're masking frustration, and whether the goal you set last week was too easy or too hard. You're adjusting on the fly based on what you see in front of you, not what a model predicts. Training caregivers, coordinating with physios and speech therapists, explaining to a parent why their child's sensory processing affects handwriting — these require physical presence, empathy, and the ability to read a room. No tool built today comes close.

According to O*NET task data, your irreplaceable tasks include planning and conducting full therapy programs across hospital, institutional, and community settings, selecting individualised activities within each patient's physical or mental limits, and evaluating patient progress over time. These aren't just hard to automate. They're hard to even define precisely enough for a model to attempt. The variability in human bodies, human behaviour, and human context is exactly what makes this work resistant to any system that relies on pattern-matching.

The one task in your role that AI can assist with is consultation and education to groups like parents and teachers. Even there, the penetration rate sits below 85%, meaning it's a partial assist at best. You might use AI to draft a handout or summarise research on sensory processing. But the conversation itself, reading whether a parent is overwhelmed, whether a teacher is actually going to follow through, whether the room needs a different approach — that stays with you.

AI also can't be held accountable. When you document a patient's progress and make a clinical recommendation, your name and licence are on it. AI-generated clinical notes in healthcare have been shown to introduce factual errors and omit context that changes the meaning of a patient's condition. In a field where your documentation directly informs a patient's care plan, insurance coverage, and legal record, you can't hand that to a system that hallucinates details.

There's also the physical environment. You lay out materials, you observe how someone picks up scissors, you clear and reset a therapy space. AI has no presence in that room. It can't see that a patient's left hand is compensating for their right, or that they're rushing through an activity because they're embarrassed. That observational layer is where most of your clinical insight comes from.

AI's footprint in occupational therapy is genuinely small. The only task with any measurable AI penetration is group consultation and education — and even there, the tools are assistive, not autonomous. You might use an AI writing tool to draft parent education materials faster, or use something like ChatGPT to pull together a plain-language summary of sensory integration research for a teacher. That's about as far as it goes for direct clinical tasks.

On the documentation side, tools like DAX Copilot and Nabla can transcribe and draft clinical notes from voice recordings. A therapist who spends 45 minutes at the end of the day typing progress notes can cut that to under 10 minutes with these tools. They're not OT-specific, but they work across allied health settings. The notes still need your review — and as covered above, you can't take that review lightly — but the time saving is real.

For home modification assessments, tools like Occupational Therapy Assistant AI (a newer category of planning aids) can generate preliminary checklists for fall risk and accessibility based on patient data. These are useful starting points. They don't replace the site visit or your clinical eye, but they can reduce the time you spend building forms from scratch. Some OT practices are also using general scheduling and caseload management platforms with AI features, like Jane App, to reduce admin overhead. None of these touch your clinical work. They sit around the edges of your day.

The biggest shift isn't in how you treat patients. It's in the time you spend before and after. If you're using documentation tools, you're probably spending less time on end-of-day note-writing and more time reviewing AI-drafted text for accuracy. That review step is non-negotiable, but it's faster than writing from scratch. The rhythm of your admin load has shifted from writing to checking.

What hasn't changed is everything that happens in the room. Assessment, hands-on intervention, caregiver training, team coordination — these run exactly as they always have. You're not spending less time with patients. If anything, some OTs report spending slightly more direct contact time because the documentation tools covered above are handling the first draft of their notes.

The group education part of your work — the parent sessions, the teacher consultations — is the one area where some OTs are using AI-drafted materials to prep faster. A session outline that used to take 30 minutes to put together might take 10. But you're still the one running the session, reading the room, and adapting when someone in the group raises something the outline didn't cover.

The BLS projects 13.8% growth for occupational therapists between 2024 and 2034. That's nearly double the average growth rate for all occupations, which sits around 7%. With 160,000 OTs currently employed and 10,200 openings expected annually, demand is outpacing supply in many regions, particularly in rural areas and schools.

This growth isn't driven by AI filling gaps. It's driven by an ageing population, expanded school-based services, and growing recognition of OT's role in mental health and paediatric development. The patients who need OT are getting older, more numerous, and more complex. AI isn't going to provide a home assessment for an 85-year-old after a fall. You are.

According to BLS occupational projections, the demand increase is concentrated in outpatient care centres, home health settings, and schools — all environments where the hands-on, observational, and relationship-based work is most intense. These aren't settings where AI tools are expanding to take on clinical tasks. They're settings where more OTs are needed, full stop.

Given the 0.008 AI exposure score, this role's automation risk is about as low as it gets in healthcare. For that to change significantly, AI would need to develop reliable physical presence through robotics, accurate real-time behavioural assessment from video, and the ability to form therapeutic relationships. None of those are close. Robotics in rehabilitation exists (tools like Hocoma's Armeo are used in some neuro rehab settings) but they operate under an OT's supervision, not instead of one.

In the next five years, expect the documentation and admin tools to get better and more OT-specific. Expect more AI-assisted home modification software and maybe some triage tools that help you prioritise your caseload. But the exposure score is unlikely to rise above 10-15% in that window. The 10-year picture looks similar. The tasks that make up this role are defined by their resistance to standardisation, and standardisation is what AI needs to work.

Your safest move is to go deeper into the tasks that have zero AI penetration. That means getting stronger at complex case formulation, multi-disciplinary team coordination, and caregiver training. These aren't just safe from automation — they're the tasks that define senior OT practice and command higher pay. An OT who's exceptional at translating a patient's functional picture into a clear, actionable plan for a whole care team is not a role that gets automated or deprioritised.

Specialisation matters here. The OTs with the most job security are those working in settings where the human layer is thickest: paediatric sensory processing, complex neurological rehab, mental health OT, and home-based care for older adults. If you're early in your career, choosing a specialisation in one of these areas gives you a patient population whose needs are deeply individual and relationship-dependent.

Learn how to use the documentation tools well. Not because your job depends on it, but because an extra 30-40 minutes a day adds up to real time you can put back into patient contact or professional development. Beyond that, the honest advice is this: occupational therapy is a profession built on the idea that meaningful activity shapes human health. That idea doesn't get automated. Your job is to keep being genuinely good at it.