AI Robotic System Unveiled as US Faces 8% Physical Therapist Shortfall

On October 20, 2026, Johannes Lachner, an assistant professor at Purdue University and former MIT postdoctoral fellow, and Noah Geiger, a former MIT visiting student, unveiled a robotic therapy system that employs generative artificial intelligence and real-time force feedback to learn from human therapists[5]. This development emerges as the United States faces a projected 8.2% shortfall in physical therapists by 2027[3][15].

This technology directly addresses a significant demand in stroke rehabilitation. Millions globally suffer strokes each year, leaving about 5 million individuals with lasting disabilities[1]. Their recovery is often hampered by limited access to consistent care due to a growing scarcity of physical therapists[1]. For example, UK hospital stroke services are 15% understaffed, and community services experience even greater deficits[2]. In the U.S., a national shortfall of 12,070 physical therapists was reported in 2022, representing 5.2% fewer than needed[3][14]. Many stroke patients in the U.S. receive minimal therapy, with some getting only six to eight sessions within three months post-stroke and very few sessions for the rest of the year[3]. Steven Cramer, lead author of a February 2023 UCLA-led study, observed that “too many patients are missing out on a golden opportunity to maximize recovery during a critical period following a stroke”[3].

The Shortage of Therapists

The American Physical Therapy Association's (APTA) workforce model, released in March 2025, confirmed a national shortfall of physical therapists in the U.S., projecting that demand will continue to exceed supply[13][14][16]. This shortage is forecast to reach 8.2% by 2027, an increase from 5.2% in 2022[15]. A survey from December 2024/January 2025 in the UK, conducted by the Chartered Society of Physiotherapy (CSP) and the Association of Chartered Physiotherapists in Neurology (ACPIN), found hospital stroke services to be 15% understaffed, with community services facing larger deficits[2]. Patients in the UK receive rehabilitation only three to four days a week in hospitals and one to two days after discharge, which is less than the recommended three hours a day, five days a week[2].

AI-Powered Robotic Therapy

AI-enhanced robotic therapy tackles these challenges by providing intensive, repetitive, and task-specific training, which is essential for neuroplasticity—the brain's ability to reorganize itself[4][12]. AI integration supports personalized, adaptive, and data-driven interventions, allowing for real-time monitoring and feedback during rehabilitation[4]. The system developed by Lachner and Geiger allows human therapists to train the AI by wearing force-sensing gloves during patient treatment[5]. This process enables the AI to infer clinical responses and dynamically adjust assistance, delivering what Lachner described as “just the right amount” of help[5]. He stated that the goal is “not to replace therapists, but to extend their reach”[5][19].

Evidence of Effectiveness

Growing clinical evidence supports the efficacy of AI-driven robotic therapy. A 2026 systematic review noted the increasing adoption of AI, virtual reality, gamification, and telerehabilitation to enhance training intensity, engagement, accessibility, and real-time monitoring in neurological rehabilitation[6]. While outcomes are often comparable to conventional therapy, some studies indicate superior results[6]. A June 2025 meta-analysis by Alsaeed et al. concluded that AI and robotics-based rehabilitation significantly improve motor and functional recovery post-stroke, showing improvements in motor function and balance[7]. A 2023 study by Khalid et al. also showed that AI-driven robotic exoskeletons could reduce rehabilitation time for post-stroke patients by 30% compared to conventional therapies[7][21].

Obstacles to Widespread Use

Despite the promising benefits, barriers to widespread adoption remain. Robotic rehabilitation equipment can be expensive, potentially costing “up to several hundred thousand dollars per device”[8]. While some studies suggest robotic training can be more cost-effective per patient session or increase therapist efficiency, leading to an overall reduction in cost per patient, findings on cost-effectiveness are mixed[9]. A NewYork-Presbyterian study in November 2024 found that only 20% of clinicians had used robotic devices, though the majority (64.9%) perceived them as useful[9]. Clinicians generally hold positive views on the therapeutic use of robotic technology, recognizing benefits such as patient engagement and objective progress evaluations, but they also cite practical implementation challenges, including lack of awareness and technical difficulties[9]. Weihua Song, a researcher in the field, stated in October 2026 that AI should serve as an “adjunct to multidisciplinary care rather than a replacement for clinicians”[10]. Ongoing research, particularly large-scale clinical trials, is needed to further validate efficacy and optimize implementation[11].

What we could not confirm

We could not confirm the specific cost-effectiveness of the newly unveiled AI robotic therapy system relative to existing robotic solutions[8][9]. We also could not confirm the specific technical difficulties and capital outlays this particular system will face for widespread adoption in clinical settings[8][9].