Will AI Steal My Job? · Role analysis

Physiotherapist

O*NET 29-1123.00 ESCO: Physiotherapists

Changing

Physiotherapists assess, diagnose, and treat physical conditions affecting movement and function through exercise, manual therapy, and rehabilitation. They work with patients recovering from surgery, injury, stroke, or managing chronic conditions — combining hands-on treatment with education and exercise prescription.

Section 01

Task Map

Task	AI impact	Why
Conduct physical assessments and movement analysis	🟡 Changing	AI motion analysis tools (using cameras or wearables) can quantify movement data, but the clinical interpretation still requires expert judgment.
Deliver hands-on manual therapy	🟢 Safe	Joint mobilisation, soft tissue release, and manipulation require trained hands with continuous tactile feedback. No robot yet matches this fidelity in clinical settings.
Design personalised exercise rehabilitation programmes	🟡 Changing	AI apps can generate exercise plans from symptom input, but tailoring rehabilitation to an individual's goals, pain tolerance, and home situation needs clinical skill.
Supervise exercise sessions and correct technique	🟢 Safe	Real-time observation of movement quality, verbal cueing, and motivational coaching during exercise is a physical, interactive skill.
Write treatment notes and progress reports	🔴 High exposure	Structured clinical documentation is AI-automatable. Voice-to-text with clinical NLP tools can draft notes from spoken summaries after sessions.
Educate patients on self-management strategies	🟡 Changing	Patient education apps can deliver information, but adapting advice to a patient's health literacy, fears, and lifestyle in conversation is human work.
Liaise with medical teams on patient progress	🟢 Safe	MDT communication and clinical advocacy — arguing for a patient's rehabilitation needs in a ward round — requires professional presence and relationships.
Assess and manage falls risk in elderly patients	🟡 Changing	Sensor-based falls risk tools are improving, but the clinical conversation and trust-building with an elderly patient about changing their environment and habits is human.

Section 02

What Stays Human

Manual therapy and therapeutic touch The therapeutic relationship established through physical contact — knowing how hard to push, when to back off, and how to read a patient's pain response in real time — is a skill built over years of practice. No current robot delivers this.
Motivating patients through painful recovery Rehabilitation often hurts, is slow, and requires the patient to keep working when they want to give up. The physiotherapist's role in sustaining that motivation, understanding the patient's fears, and adjusting the approach to the individual is fundamentally relational.
Clinical reasoning in complex presentations When a patient's pain doesn't fit the expected pattern — when it may be biopsychosocial, referred from another structure, or masking a serious condition — the diagnostic reasoning required draws on years of pattern recognition that AI tools cannot yet replicate reliably.

Section 03

What to Do Next

Explore digital physiotherapy platforms like Kaia Health, Hinge Health, or PhysioTrack. Understanding how these work — and where they fall short — means you can use them as adjuncts to your practice rather than feeling threatened by them. Many physios now use them to scale home exercise compliance.
Specialise in a higher-complexity area: MSK extended scope practice, neurological rehabilitation, pelvic health, or sports performance. The more complex the clinical reasoning required, the more protected the role from automation. Extended scope practitioners with prescribing rights are particularly in demand.
Develop your patient education and self-management programme skills. As routine exercise delivery becomes increasingly app-supported, the physiotherapist who designs the programme — and handles the complex cases where the app isn't enough — becomes the value-added clinician.

Sources: O*NET Online (onetonline.org) · ESCO (esco.ec.europa.eu) · All task data cross-referenced against O*NET occupation profiles. This analysis uses task-level exposure, not occupation-level prediction.