When adherence breaks on language
With patients who don't share your language, the session itself usually works. You demonstrate the exercise, the patient copies it, and what you can't say you show. The harder part is what goes home: sets, reps, breathing, what to do if it hurts. That's on paper or in a link, and usually in your language, not theirs. This piece is about that part, and what AI translation does and doesn't change about it.
How often this happens
The numbers below come from the Netherlands, where Kinmo is based. Almost 2.5 million people there don't have Dutch as their first language (Pharos, 2025). In 2022 Berenschot surveyed 466 healthcare providers across sectors about interpreter use, for the Dutch Ministry of Health. 15% run into a language barrier often or very often. In the situations where providers themselves felt an interpreter was needed, one was rarely or never brought in 72% of the time (43% never, 29% rarely).
What it costs
When a patient didn't understand the instructions, you usually don't see it as that. You see a patient who isn't doing the exercises, or is doing them wrong. So you re-explain what you already explained, or you scale back a program that wasn't too hard in the first place, and sometimes the patient drops out.
Adherence to home exercise is shaky to begin with. In their overview of systematic reviews, Ley & Putz (2024) note that non-adherence can run as high as 70%, particularly with unsupervised home programs. How much a language barrier adds to that hasn't been well studied for home exercise. That it affects the care around it is clear. A systematic review of eight studies on how physical therapists perceive the quality of their own care found that language barriers hamper, among other things, history-taking, the use of standardized tests and building rapport with the patient (Neipert et al., 2024). In US orthopedic care, a non-English language preference is consistently associated with worse postoperative outcomes and lower patient-reported outcome scores (Garcia-Lopez et al., 2025).
How practices handle it today
There are roughly three routes, and each leaves something open.
Phone or video interpreter. Works during the session, but the patient goes home without one. The written part, with what she needs to do over the next week, stays in your language.
A family member as interpreter. Quick and free, but privacy isn't covered, children sometimes interpret for their parents where they shouldn't, and you don't know whether that relative is around when the patient exercises at home.
Google Translate. Easy to reach for, but it has no clinical context. A translation app doesn't know that "bridge" or "clamshell" is an exercise and translates it as an ordinary word. "Hold for 10 seconds" comes out grammatically fine, but not always in a way that makes clear what is being held. And you can't check what comes out the other end in a language you don't read.
What AI translation in the exercise program solves
If translation is built into your exercise program tool, it covers part of what the other routes leave open:

- The translation belongs to the program. The patient opens it at home and reads the instructions in Polish, Turkish, Arabic or whatever language she speaks.
- The source text is always the same. If you work from a fixed exercise library, every translation starts from the same name and the same description of how to do it, not from whatever you typed into a translation app that day.
- You keep working in your own language. There's no bilingual program to build.
At home, the patient then reads what was discussed in the session, in a language she actually knows.
What it doesn't solve
Translation doesn't replace the session. Examination, explanation and demonstration still happen in the treatment room, in whatever language you find together. For the history or a difficult conversation, a professional interpreter is still the better choice.
Medical terminology remains a risk. Language models have improved, but they don't always get clinical jargon right, and a term that is precise in one language can mean something slightly different in another. You can't proofread an AI translation in a language you don't read either. For instructions that matter, such as contraindications and red flags, check in the session (with an interpreter if needed) that the patient understands the point.
Translation also does nothing about cultural differences: how someone views pain, movement, exercise or the role of the clinician. Those are separate conversations.
What this means for your practice
If you regularly see patients who don't share your language and adherence lags in that group, look at where it goes wrong. For the session, an interpreter is the answer. For the time between appointments, when the patient exercises alone, translated written instructions help.
Where Kinmo fits
Kinmo is our own product. You build the program in your language. You pick a language, or the patient does when she opens the program, and Kinmo uses AI to translate the whole program, including your own exercises and notes, into any of more than a hundred languages. You can do that up to five times per program. The patient doesn't need an app or an account.
Beyond that, Kinmo only uses AI when you import exercises or a protocol: it suggests missing names and fields, which you review before they go into your library. Building and sending a program involves no AI.
Sources
- Ley, C., & Putz, P. (2024). Efficacy of interventions and techniques on adherence to physiotherapy in adults: an overview of systematic reviews and panoramic meta-analysis. Systematic Reviews, 13, 137. doi:10.1186/s13643-024-02538-9
- Neipert, A., Russo, K., Ortt, S., et al. (2024). Impacts of language barriers on perceived quality of care in physical therapy: a systematic review. Bulletin of Faculty of Physical Therapy, 29, 51. doi:10.1186/s43161-024-00216-4
- Berenschot (2022). Tolken in de zorg: een overzicht van huidige inzet, financiering en knelpunten (n=466). Commissioned by the Dutch Ministry of Health, Welfare and Sport.
- Pharos (2025). Generieke richtlijn Omgaan met taalbarrières in de zorg en het sociaal domein. pharos.nl
- Garcia-Lopez, E., O'Marr, J., Gottlieb, R., Miclau, K.R. & Pandya, N. (2025). Language Barriers in the Delivery of Musculoskeletal Care and Future Directions. Current Reviews in Musculoskeletal Medicine, 18, 559-567. doi:10.1007/s12178-025-09986-3