Will AI take over therapist jobs? See what Hinge Health and Sword Health's AI rehab platforms automate in 2026, and what still needs a licensed therapist.
Our AI Job Risk Index currently scores Therapist at 11 out of 100. A higher score means more of the role's routine, well-defined tasks can already be automated — it is not a prediction that the profession disappears. AI tends to absorb repetitive work first, while judgement, accountability, and human relationships stay with people.
Therapists, physical, occupational, and rehabilitation therapists in particular, do a great deal more than administer pre-set exercises. Their work is to adjust the amount, order, and form of support to fit a person's physical condition, recovery stage, pain, anxiety, and life goals. The role involves not just delivering techniques but designing support that the person can actually continue at home.
AI can streamline the surrounding work, documentation, standard-program suggestions, and remote movement tracking, but the intervention itself remains strongly human. The therapist's value lies in reading the person's response in the moment and shaping support around it, which is exactly the part that digital rehab platforms still route back to a licensed clinician.
April 1, 2026
Stanford’s findings on the dangers of chatbot personal advice weaken the argument that conversational AI can safely replace therapeutic work in the near term. Since therapists rely on trust, nuanced judgment, and safeguarding vulnerable clients, this week’s evidence supports a small downward revision in AI replacement risk.
March 18, 2026
Safety concerns around AI chatbots in mental-health settings, including reports of psychosis and severe harm, reduce confidence in direct AI substitution for therapeutic care. Because therapist work requires trust, accountability, and intervention in vulnerable situations, the risk score moves down slightly from last week.
AI will not replace therapists, but it is already automating a meaningful share of what used to require a clinic visit. Hinge Health, a digital musculoskeletal-care company that went public on the NYSE in May 2025, says its AI-powered motion-tracking technology, TrueMotion, cuts the human care-team hours needed per patient by around 97 percent compared with in-person physical therapy, mostly by letting a camera and algorithm supervise routine home exercises that used to require a therapist watching in the room. That is a real shift in who delivers the repetitive part of a rehab program. It is not evidence that therapists are becoming unnecessary.
The clearest sign of where the line actually falls is Sword Health, a roughly $3 billion digital-therapy company that cut 13 physical therapist positions in 2025 as it shifted toward having AI handle more of each remaining clinician's caseload, each human therapist overseeing more AI-guided patients rather than being replaced outright. Sword also acquired the UK prehabilitation company Surgery Hero in 2025 and rolled its programs out across 18 NHS trusts, expanding how much pre-surgical and standard-program work AI can carry before a person needs to step in.
What both companies still route to a licensed therapist is the part that doesn't standardize: reading pain, fear, and fatigue in real time, deciding when to push and when to stop, and rebuilding a program around a person's actual daily life rather than a generic protocol. Therapists do more than provide techniques. They design support that people can actually continue in real life. A better way to look at the role is to separate the tasks likely to be accelerated by AI from the parts of therapy that remain human at their core.
AI fits best into therapy tasks built around documentation, standard options, and remote monitoring. Hinge Health's own reported numbers show how far that can go for the repetitive, well-defined part of a rehab program, and how much of the caseload growth at companies like Sword Health depends on AI carrying that load instead of a person.
AI can produce rough drafts of standard therapy programs based on goals and common conditions, and platforms built on computer-vision motion tracking can now supervise a patient running through those exercises at home without a therapist in the room. That is useful as a starting point and for routine maintenance work, but deciding whether the program truly fits the person, and adjusting it when it doesn't, still depends on professional judgment.
AI can organize first drafts of therapy records and progress summaries efficiently, which is part of why companies scaling AI-guided care can grow patient volume per clinician so sharply. Even so, therapists still need to decide what parts of the response are clinically meaningful rather than just recorded.
Sensor- and camera-based tracking can now count repetitions, check form against a reference movement, and flag when someone isn't completing a program as prescribed. That kind of monitoring used to require a scheduled clinic visit. It works well for maintenance-phase exercises with a known correct form, and far less well for early-stage recovery where the correct movement is still being figured out.
AI can assist in organizing evaluation metrics and making progress visible across a caseload, part of how digital-health platforms report outcomes to the employers and health plans that pay for them. Deciding what should be emphasized next, and how those results should change a specific person's support, still remains human.
What remains with therapists is the work of adjusting support based on real-time human response and designing support around a person's life goals. The more a task depends on immediate feedback, motivation, and judgment about safety, the more strongly it remains human, which is exactly why even the most aggressive digital-rehab platforms keep a licensed clinician in the loop rather than removing one.
Therapists still need to judge how far to push, when to stop, and how to change support based on pain, fear, fatigue, or emotional response in the moment. A camera can confirm that a movement matches a reference pattern; it cannot judge whether today is a day to push through discomfort or back off.
Therapists still need to help people sustain motivation through difficult recovery periods, speaking in a way that gives realistic hope without ignoring difficulty. This relational work is precisely why digital-rehab companies describe a human clinician as still overseeing every AI-guided program, not just the complicated ones.
Therapy still has to connect to a person's actual life, work, home, recovery priorities, not just to a function score on a dashboard. The work of aligning intervention with daily living remains a core professional role that a standard program, however well it tracks motion, cannot substitute for.
Therapists still need to coordinate with doctors, nurses, families, and other support workers so progress continues across settings. That bridging role sits outside what any single monitoring platform captures, and it remains difficult to automate.
For therapists, future value depends less on standard-program knowledge alone and more on the ability to observe responses, adjust support, and connect therapy to real-life recovery. As AI absorbs more of the routine monitoring, the individualized judgment layer on top of it becomes the whole job.
Therapists need to do more than notice reactions; they need to interpret what those reactions mean for the next step in support. As AI increasingly handles the record-keeping and rep-counting, this interpretive ability is what separates a therapist's caseload from a monitoring subscription.
Good therapy goes beyond isolated exercises. It is about helping people rebuild function in the context of daily living. Therapists who think from that broader perspective will remain stronger as more of the exercise-tracking layer gets automated around them.
Communicating in a way that makes continued effort realistic rather than overwhelming matters more once an app is also nudging the patient daily. Therapists who can turn difficult support into something people can actually keep doing remain highly valuable.
Even when an algorithm produces a clean standard plan or motion-tracking target, therapists still need to break from it when a person's condition calls for a different path. The more capable AI becomes at producing generic plans, the more valuable it is to know when to abandon them.
Therapy experience builds strengths in observation, individualized adjustment, motivation support, and recovery design. That makes it easier to move into nearby roles where human support and practical judgment both matter.
Experience in close, ongoing support and functional recovery translates well into nursing roles focused on everyday patient care and monitoring recovery outside a single session.
Therapists who already work through dialogue and behavioral change may adapt well to counseling-centered roles, where reading emotional response and adjusting the conversation in real time is the whole job rather than one part of it.
Experience connecting support to real-life conditions also supports social-work roles focused on continuity, systems, and daily living, especially for patients whose recovery depends on circumstances outside the clinic.
Therapists who enjoy adapting a program to what someone can actually do at home and work often move naturally into occupational therapy, which shares the same emphasis on individualized, real-life-focused adjustment.
People who want to expand from intervention support into broader treatment responsibility may move toward physician roles, building on the diagnostic instincts therapy already requires.
Experience helping people recover capacity and stay engaged connects naturally to counseling roles focused on returning to work and rebuilding direction after a disruption.
Therapists are not becoming unnecessary. What's changing is who does the repetitive part of the job: platforms like Hinge Health and Sword Health are demonstrating, with real patient volume and public financial results, that AI-guided monitoring can absorb a large share of routine exercise supervision. What remains is the work of adjusting support based on real-time response, sustaining motivation, linking therapy to life goals, and coordinating long-term care across providers. As the work changes, career strength will depend less on standardization and more on individualized human judgment.
These roles appear in the same industry as Therapist. They are not the exact same job, but they make it easier to compare AI exposure and career proximity.
Our AI Job Risk Index currently scores Therapist at 11 out of 100. A higher score means more of the role's routine, well-defined tasks can already be automated — it is not a prediction that the profession disappears. AI tends to absorb repetitive work first, while judgement, accountability, and human relationships stay with people.
The score combines a baseline estimate of how automatable the role's core tasks are with a weekly re-evaluation that weighs the latest AI research, products, and news. Scores are relative across every tracked job, so Therapist's number is best read in comparison with other roles rather than as an absolute probability.
No role is fully insulated, but you lower your exposure by leaning into what AI handles worst: complex judgement, ethical accountability, hands-on or interpersonal work, and supervising AI output. Workers who use AI as a tool consistently fare better than those who try to compete with it.
The score is updated every week from our index. The weekly-change figure on this page shows how much Therapist's AI exposure shifted compared with the previous week.
