This study evaluated the interobserver reliability of intraoperative soft tissue laxity assessment during robotic-arm-assisted total knee arthroplasty (TKA) using the Robotic Surgical Assistant (ROSA) system. Although robotic systems have improved the precision of bone alignment, the reproducibility of soft tissue evaluations—critical for achieving joint balance—remains uncertain. We retrospectively analyzed 37 knees from 30 patients who underwent ROSA-assisted TKA. Soft tissue laxity was manually assessed under varus and valgus stress at extension and 90 degrees of knee flexion. Two parameters were recorded: joint gap widening and hip–knee angle (HKA). Interobserver agreement was evaluated using intraclass correlation coefficients (ICC), Pearson's correlation coefficients, and Bland–Altman analysis. Angular measurements, particularly HKA at extension, demonstrated excellent interobserver reliability (ICC = 0.94). In contrast, gap-based measurements showed greater variability, with lateral laxity under varus stress at extension demonstrating poor reliability (ICC = 0.31). Agreement between observers was generally stronger at extension than at 90 degrees of flexion. Contralateral assessments showed better consistency. While robotic systems improve alignment accuracy in TKA, intraoperative soft tissue laxity evaluation remains prone to interobserver variability. Gap-based measurements are less reliable than angular assessments, underscoring the need for standardized evaluation methods to improve consistency in soft tissue evaluation during robotic TKA.