Why the airway stays with a person
Respiratory therapy is bedside work with a narrow margin for error. A therapist sets up and adjusts mechanical ventilators, suctions airways, and reads how a patient actually responds in the next thirty seconds. Software can suggest a setting. It cannot hold a mask on a frightened patient, feel resistance in a suction catheter, or decide in a crowded room that the plan needs to change.
The physical share of the job is the main reason it moves slowly. Our robotics read for this work puts it in the dexterous humanoid tier: to take over the hands-on tasks, a machine would need steady hands near a face and an airway, in an unstructured hospital room, with no room for a dropped step. Nothing at that level is working in US hospitals at scale.
There is also accountability. Therapists draw and analyze arterial blood samples, respond to codes, and teach patients and families how to use inhalers and home oxygen. Those tasks carry a licensed signature and a conversation. Both are hard to hand to a tool. You can see how the whole picture is weighted in how we score jobs.
What AI does, assists with, and leaves alone
Start with the work people keep. The share of task time our scoring leaves with a human is 80% of the job. That is where airway management, ventilator weaning at the bedside, emergency response, and patient coaching sit. The task list above shows which items land in that group.
Next, the part AI can run with little supervision: 0% of task time. These are the paperwork-shaped tasks — charting treatments, pulling monitoring data into a readable summary, flagging alarm patterns across a unit. The output still gets checked by the therapist who signs it.
Then the assisted middle, at 20% of task time. Here the tool drafts and the person decides: suggested ventilator settings, a first pass at pulmonary function test interpretation, sleep study scoring sent for review. Overall coverage — how much of the job AI could handle today — reads 12 out of 100. What that number counts is explained on the coverage method page.
What the evidence actually shows
Evidence grade for comparing AI against a therapist is D. That is our lowest grade, and it means something specific: there is no published head-to-head test of an AI system against a qualified respiratory therapist doing this job. So we publish no parity number for it. Claims that a model matches or beats a therapist are not measurable from what exists.
Plenty of research looks at pieces of respiratory care — image reading, spirometry interpretation, alarm prediction — but a task-level result is not a job-level result. A grade would move if someone ran a prospective study in a real ICU: the same patients, the same ventilator decisions, an AI-guided arm and a therapist-guided arm, with outcomes like ventilator days, reintubation and complications tracked. Until that exists, the honest read is task erosion in documentation and interpretation, not a job handed over. The sources we do use are listed at the foot of this page.
What could move the date, and what holds it
Most likely after 2042 (8 in 10 of our scenarios). For what that window measures and how it is built, see the replacement-year method.
Two things could pull it earlier. First, closed-loop ventilation: if adaptive modes prove themselves in trials and regulators clear them for broad use, a slice of titration work moves off the therapist’s hands. Second, cost. Running software on the interpretive and charting tasks is cheap next to staffing them, and the cost panel above shows how wide that gap is. Hospitals under budget pressure act on gaps like that, usually by thinning entry-level hours first.
Two things hold it back. The physical majority of the work still needs hands in the room, and the robots that could do it are not deployed. And liability sits with a licensed clinician: when a vent change goes wrong, someone has to answer for it. Hospital capital cycles are slow too, which is why the hospitals sector page moves less sharply than office-heavy sectors.
Good to know: BLS projects employment for respiratory therapists to grow about 8.5% between 2025 and 2035, against median pay of $82,280 and about 139,790 US jobs (BLS, 2025).
How to stay needed in respiratory care
Lean into the tasks that keep the job yours. Three worth building on: airway and emergency work, where you are the person called; bedside ventilator management and weaning, where judgment beats a default setting; and patient and family education on inhalers, home oxygen and airway clearance, which is teaching, not output.
Two skills pay off alongside that. One is reading machine output critically — knowing when a suggested setting or an auto-scored study is wrong and being able to say why in a chart note. The other is cross-unit communication: NICU, ICU, ED and transport teams all rely on a therapist who can hand off cleanly under pressure.
If you are weighing a move, nearby work includes Registered Nurses, Cardiovascular Technologists and Technicians and Paramedics. You can put any two jobs side by side on the compare page, see the rest of the family on the healthcare practitioners and technical page, or check where hands-on clinical roles land on our list of jobs that mostly need a person. The headline figure here reads 82 out of 100 (higher is safer), and how that score is built is published in full.