<p>Ancient Chinese Character Recognition (ACCR) poses significant challenges owing to the complex structural variations of ancient Chinese characters. Traditional approaches, which primarily focus on global features, often fail to capture the intricate details and subtle structural variations, resulting in suboptimal recognition performance. To overcome this limitation, a Hierarchical Adaptive Structural Pooling Network (HASP-Net) is proposed for ancient Chinese character recognition. It optimizes two tasks jointly: character classification and spatial structural prediction. Specifically, an Adaptive Structural-Aware Pooling Aggregation module was designed, which employs pooling windows of various sizes and aspect ratios to correctly respond to various structures and primitives in high-level feature maps, thereby comprehensively capturing the structural information of Chinese characters. Furthermore, the prediction of characters’ structural information guides the model to adaptively select different pooling windows, effectively aggregating local features from different regions. Extensive experiments on ancient Chinese character dataset CASIA-AHCDB demonstrate that HASP-Net establishes new state-of-the-art results.</p>