Impact of Artificial Intelligence or Telerehabilitation on Consumer Ratings in Physical Therapy: An Online Randomized Experiment
Related Experiment Video
Updated: Nov 2, 2026
Application of Bedside Lower Extremity Rehabilitation Robots in Stroke Rehabilitation: A Randomized Controlled Trial
Published on: November 28, 2025
Megan H Ross1, Joshua Simmich1, Daniela Berg1
1RECOVER Injury Research Centre, The University of Queensland, Brisbane, Queensland, Australia.
Importance:
Artificial intelligence (AI) is increasingly integrated into health care delivery, but its influence on consumer perceptions of physical therapist services remains unclear.
Objective:
The objective was to examine how the use of AI influences consumer perceptions of in-person and telerehabilitation physical therapist services.
Design:
This study used a randomized 2 × 2 factorial experimental design comparing AI use and delivery mode.
Setting:
The experiment was conducted online in Australia.
Participants:
The study included 710 Australian adults recruited through a paid panel.
Exposure:
Participants were randomized to 1 of 4 hypothetical physical therapist consultation scenarios (in-person vs telerehabilitation; AI used vs not used).
Main Outcomes And Measures:
Perceived warmth; competence; safety; provider trust; emotional trust; relational and functional service quality; working alliance; and intention to use were measured on 7-point Likert-scales. Composite mean scores were analyzed using ordinary least squares regression with heteroskedasticity-robust standard errors.
Results:
Mean scores in the no-AI, in-person condition ranged from 4.8 to 5.8. Among in-person scenarios, described AI use was associated with lower ratings across all outcomes, including warmth (b = -0.3, 95% CI = -0.5 to -0.1), competence (-0.4, -0.6 to -0.2) and safety (-0.8, -1.1 to -0.6), provider trust (-0.5, -0.7 to -0.3) and intention to use (-0.8, -1.0 to -0.5). Without AI, telerehabilitation also reduced ratings by 0.2 to 0.7 points across outcomes. Positive interaction effects indicated attenuation of AI-related reductions in telerehabilitation for relational (interaction b = 0.4, 95% CI = 0.1 to 0.8) and functional service quality (0.36, 0.02 to 0.70), intention to use (0.6, 0.1 to 1.0), and perceived safety (0.5, 0.1 to 0.9).
Conclusions:
Hypothetical in-person consultations described as involving AI were associated with poorer consumer perceptions of physical therapists and services. Telerehabilitation was also perceived less favorably than in-person care.
Relevance:
Future studies should test whether transparency and reassurance mitigate the effects of AI-use on consumer perceptions of physical therapy.
