Why this work keeps a person in the room
Exercise physiologists test how a heart, lungs and muscles answer effort, then build and supervise the program that follows. The test is the job. Electrodes go on skin. A blood pressure cuff is read at each stage of a graded treadmill or cycle protocol. Someone watches the face, the gait and the breathing while the load climbs, and decides when to stop. Software can read a trace. Deciding to end a stage early, for this person, on this day, is a judgment call with a body in front of it.
The other half of the role is talking. Reviewing a medical history with a client, explaining why their program starts slower than they hoped, demonstrating a movement and correcting it by hand, and keeping someone coming back after week three. Adherence is not a data problem. That is the main reason the answer to the question of whether AI will replace exercise physiologists is not a clean yes.
Scale matters too. This is a small occupation: about 8,560 US jobs with median pay of $59,460, and projected employment growth of 12.8% over 2025 to 2035 (BLS). Small, growing occupations rarely attract purpose-built automation, because the market for a machine that straps on an ECG and cues a lunge is thin.
What AI does, what it helps with, what stays human
Where AI already carries work, it is the paperwork and pattern-reading around the session: pulling wearable and heart-rate data into a readable summary, drafting progress notes, flagging values outside a range, and producing a first-pass program from a template. That slice of task time shows as 0% of the job. It is real time saved, mostly the time that used to be spent typing after the client left.
A larger block is assisted rather than handed over. Interpreting test results, comparing this month’s numbers to last month’s, and tuning a prescription are all faster with a model in the loop and still signed off by the physiologist. That assisted share sits at 37%. The Can AI do it? figure behind it is 17 out of 100, where a higher number means more task time AI can handle today; how coverage is measured explains what counts.
The rest needs hands and eyes. Supervising a stress test, spotting an adverse response and stopping, placing leads correctly on an awkward torso, teaching a movement by touch, and holding a conversation that changes someone’s behavior. That group is 63% of task time, and it is why the headline Still needs a human figure lands at 78 out of 100 (higher is safer).
How strong is the evidence?
Thin, and the page says so. The Is it better than a person? grade for this job is D, our lowest evidence tier, which means no study has tested an AI system against qualified exercise physiologists on their own work. So there is no parity number here, and anyone quoting one for this job is guessing.
What would settle it is specific: a blinded comparison of AI-generated exercise prescriptions against physiologist-written ones for the same cardiac or metabolic cases, scored on safety and outcomes; and a test of whether an automated monitor calls stop points as well as a supervising clinician during graded testing. Until work like that exists, the honest read is task erosion in documentation and data handling, not a replaced role. Our quality parity method sets out what each grade requires.
When the picture could shift
Most likely after 2042 (8 in 10 of our scenarios). That timing comes from the model described in the replacement-year method, not from any single forecast.
Two things could pull it earlier. First, cheap tooling: the cost panel on this page shows AI assistance priced well below the human hours it would stand in for, which makes employers willing to try it on the admin half. Second, wearables. If continuous monitoring becomes accurate enough for clinical decisions, more assessment moves out of the lab and into an app.
Two things hold it back. The physical share of the work needs a body in the room, and the robotics panel here points to dexterous humanoid hardware rather than anything a clinic can buy and deploy this year. Clinical responsibility is the other brake. Supervised testing on patients with heart or lung conditions sits inside medical oversight, and that paperwork moves slowly.
Good to know: the biggest near-term squeeze is on entry-level hours, since data entry and note drafting were often the junior tasks.
How to stay needed
Lean into the parts of the job that stay with people. Own supervised testing on complex cases, including stop-point judgment and emergency response. Own hands-on movement correction, where a correction is felt rather than described. Own the behavior-change conversation that keeps a program going past the first setback.
Two skills are worth real practice. One is reading and challenging machine output: knowing when a wearable’s numbers or an auto-generated program are wrong for this client, and saying why in writing. The other is clinical communication with referring physicians, because the person who translates test data into a care decision is hard to route around.
If you are weighing a nearby path, three jobs share a lot of the same work: physical therapists, recreational therapists and athletic trainers. You can put any two of them side by side on our job comparison tool, read the wider diagnosing and treating practitioners family, or see how this role sits against the rest of the healthcare sector. For the wider view, there is our list of jobs that mostly need a person and the full scoring method.