Stakeholder Perspectives and Perceived Usability of an AI‐Enabled Stroke Rehabilitation Platform: A Multicentre Mixed‐Methods Focus Group Study With Stroke Survivors, Caregivers, and Clinicians

INTRODUCTION: AI-enabled digital neurorehabilitation may support home-based stroke recovery, but its adoption depends on user acceptance. Within the European AISN project (Integrating Artificial Intelligence in Stroke Neurorehabilitation), we explored how stroke survivors, caregivers and healthcare professionals perceived the usability of a preliminary version of an AI-based rehabilitation platform, and the barriers to its adoption. METHODS: We conducted a convergent parallel mixed-methods study in four European countries (Spain, Italy, France and Romania). Seventeen focus groups combined hands-on testing of the platform components (RGS ecosystem and MIMS dashboard) with semi-structured discussions. Usability was measured with the System Usability Scale (SUS), using exploratory analyses: stroke survivors and caregivers rated the RGS, whereas healthcare professionals rated the combined MIMS/RGS system. Qualitative data were analysed thematically and integrated with SUS total and item scores in a joint display. RESULTS: Seventy-one participants took part (27 stroke survivors, 17 caregivers, and 27 healthcare professionals). The mean SUS score was 68.1 ± 16.0 (grade C). Stroke survivors rated the patient interface (RGS) lower than caregivers rating the same interface (58.9 ± 16.5 vs. 75.6 ± 13.3; p = 0.002); professionals, who rated the clinician-facing components (RGS/MIMS) after a demonstration, scored 72.7 ± 12.8. Six themes emerged: ease of use and clinical utility, personalised treatment and feedback, technical reliability and trust, the digital literacy divide, the need for human mediation and workflow and socio-economic integration. Patients' concerns about reliability converged with their lower confidence in using the system (SUS item 9), whereas digital literacy, although described as a major barrier, was not associated with patients' scores. Human mediation, clinician time and reimbursement emerged as barriers not captured by the SUS. DISCUSSION: Autonomous home use does not yet appear realistic for stroke survivors without support. All stakeholder groups favoured a hybrid model combining AI-driven exercise delivery with clinician oversight. These findings are informing the refinement of the AISN platform and the design of a forthcoming randomised controlled trial. PATIENT OR PUBLIC CONTRIBUTION: Stroke survivors (n = 27), formal and informal caregivers (n = 17), and multidisciplinary healthcare professionals (n = 27) participated in this study as relevant stakeholders. Their contribution was central to the formative evaluation of the AISN platform, including hands-on testing and discussion sessions focused on the RGS ecosystem and MIMS dashboard. Patients and caregivers contributed their lived experiences, expectations, concerns, and priorities regarding AI-assisted home rehabilitation, while healthcare professionals provided their clinical expertise, including perspectives on clinical utility and workflow integration. Stakeholders were not involved in the conceptual design of the focus group (FG) study protocol, the systematic analysis of the qualitative data, or the drafting and dissemination of the manuscript. Nevertheless, their feedback served as a cornerstone of the iterative development of the AISN system. The insights and themes identified through their contributions are directly informing final technical refinements to the AISN system architecture and the protocol design for the upcoming randomised controlled trial (RCT), helping to ensure that end-user perspectives are incorporated into the final clinical solution. While the FG script was not formally piloted with a Patient and Public Involvement and Engagement (PPIE) group, it was reviewed through multidisciplinary workshops to ensure language accessibility and clinical safety. Formal PPIE was therefore not implemented as a separate research partnership activity because stakeholders were recruited primarily as participants in a formative usability evaluation, with their experiential and professional perspectives collected as study data rather than as contributions to research design, analysis, or manuscript preparation.