Abstract Background Sarcopenia is prevalent in long-term care facilities (LTCFs) and is associated with falls, functional decline, and mortality. Aims This proof-of-concept study, within the I-COUNT pilot randomized controlled trial, evaluated the feasibility and diagnostic performance of a non-invasive sarcopenia screening approach based on wearable surface electromyography (sEMG) combined with artificial intelligence. Results Eighteen older adults residing in two Italian LTCFs (6 sarcopenic, 12 controls; 33.3% men, mean age 87 ± 5.4 years) underwent sEMG during three standardised motor tasks involving lower-limb muscles. Surface-EMG signals were transformed into time-frequency scalograms (Continuous Wavelet Transform) and classified as sarcopenic versus non-sarcopenic using deep transfer learning. The classification framework achieved 91% accuracy using Leave-One-Out Cross-Validation, with sensitivity of 84%, specificity of 93%, precision of 92%, and an F1-score of 91. Conclusions These findings support wearable sEMG as a scalable tool for screening of sarcopenia in LTCFs and provide preliminary evidence for further validation in larger populations. Trial registration number ClinicalTrials.gov ID NCT06820710.