Bipolar response has emerged in recent years as a novel operational mechanism for photodetectors. By producing switchable positive and negative photocurrents under different external conditions, it enables multidimensional mapping and multiplexing of optical signals, supporting applications such as neuromorphic vision and on-chip computing. Two-dimensional materials have seen rapid development in the field of bipolar photoresponse in recent years due to their unique optoelectronic properties, including atomic-scale thickness, tunable band structures, and strong light–matter interactions. In detector architectures based on two-dimensional materials and their van der Waals heterostructures, various types of bipolar behaviors can be engineered through external stimuli including electric field, wavelength, polarization, and incident power. These functionalities have been widely employed in brain-inspired vision, convolutional pre-processing, event-based imaging, and multidimensional spectral analysis. This review focuses on the development and applications of bipolar photoresponse in photodetectors, providing a systematic summary of its physical mechanisms, material systems, device architectures, and representative advancements. Beyond enhancing device performance, bipolar photoresponse endows photodetectors with intrinsic computational and learning capabilities, enabling functions such as logic operations, feature extraction, and adaptive perception at the device level. Furthermore, the integration of bipolar-response devices with large-scale arrays and neuromorphic hardware platforms is expected to significantly reduce system complexity and power consumption. Despite the remaining challenges in device uniformity, stability, and large-area integration, bipolar photoresponse offers new pathways for multidimensional optical information fusion, low-power visual computing, and the evolution of next-generation intelligent optoelectronic systems.