<p>In recent years, with the advancement of high-quality manufacturing, traditional teaching-based and gantry-type robotic manipulators have proven inadequate for welding tasks involving small-batch, non-standardized, or complex workpieces. Trajectory planning for humanoid welding robots (HWR) has therefore emerged as a core research direction in intelligent welding manufacturing, playing a crucial role in enhancing operational accuracy, flexibility, and intelligence, while holding significant promise for the next generation of smart welding environments. Although extensive research has been conducted on industrial robots and welding automation, systematic reviews focusing specifically on HWR trajectory planning remain scarce. This paper comprehensively outlines trajectory planning techniques suitable for HWR, summarizing their control as an integrated architecture of perception-planning-execution, and discussing trajectory planning methods in detail from these three levels. It further explores trajectory optimization under system constraints in dynamic environments, highlighting the potential of these methods to achieve efficient and stable performance in complex welding tasks. Furthermore, this paper discusses trajectory coordination problems in dual-arm collaborative welding and mobile operations, emphasizing their importance in improving the adaptability and intelligence of robotic welding. Finally, the paper summarizes current challenges and outlines future directions, including multimodal perception fusion, the application of intelligent optimization algorithms, and the integration of trajectory planning with welding process knowledge. This review aims to provide both theoretical insights and practical references for advancing trajectory planning research in HWR.</p>