This study examined the impact of augmented reality (AR) on engineering education, focusing on spatial visualisation skills and cognitive load in an engineering drawing course. The research is based on cognitive load theory and spatial visualisation frameworks. It compares three AR methods--marker-based (MBAR), markerless (MLAR) and Web-based (WBAR)--with traditional instruction. The goal was to assess how these approaches improve spatial visualisation, measured by the Purdue Spatial Visualization Test: Rotations, and reduce cognitive load, evaluated through a validated questionnaire. The results show that the MBAR group significantly improved in spatial ability and experienced less cognitive load than the MLAR and WBAR groups. This suggests that marker-based AR is particularly effective in enhancing learning, not just in engineering but also in other fields requiring strong spatial skills. The study advocates for the broader use of AR to improve cognitive efficiency and learning outcomes.