The shoulder may be affected in a large portion of patients with rheumatoid arthritis (RA) worldwide. However, this joint does not receive the attention required during follow-up. Indeed, although numerous clinical tests for diagnosis of a painful shoulder are available, differentiating articular from peri-articular lesions may be difficult in daily practice. Fortunately, the precise diagnosis of shoulder pain in RA has benefited from a reliable imaging modality used to detect its exact origin-ultrasonography (US). This study was aimed at assessing the intra- and inter-observer reliability of ultrasonographic findings for patients with established RA with shoulder pain in a patient-based exercise as a clinical challenge among Maghrebian rheumatologist experts in US. A total of 7 operators examined 10 patients in two rounds independently and blindly of each other. Before beginning the session, all of the rheumatologists reached a consensus on sites and US settings by performing a brief exercise on a normal shoulder. Outcome Measures in Rheumatology Clinical Trials (OMERACT) definitions of US-detected pathologies were used. Each patient underwent US scanning of the painful shoulder in predefined sites based on US technical guidelines of the European Society of Musculoskeletal Radiology: long head of biceps (LHB), subscapularis recess, posterior recess and axillary recess. The presence of subdeltoid or subcoracoid bursitis or full rupture of the suprasupinatus was identified if present. Intra- and inter-observer reliability measures were calculated using the κ coefficient. Intra-observer reliability was good for gray-scale (GS) synovitis in subscapularis and posterior recesses (κ=0.77 and 0.73, respectively). It was moderate in the presence of GS synovitis and effusion in LHB (κ =0.53 and 0.40, respectively), posterior and subscapularis recess effusion (κ=0.56 and 0.60, respectively) and GS and power Doppler (PD) synovitis in axillary recesses (κ=0.58 and 0.49, respectively). Inter-observer reliability was good for PD for LHB signals (κ=0.78). It was moderate for GS for LHB synovitis (κ=0.54). Inter-observer agreement was poor for effusion and GS synovitis for subscapularis, posterior and axillary recesses, and very poor for PD signals in these recesses. US was a reliable imaging tool for detecting tenosynovitis in the LHB. However, reliability was moderate to poor in detecting synovitis in subscapularis, posterior and axillary recesses. These findings could be optimized by standardization of sites to assess.