With its flexible data perception method, mobile crowd sensing technology has become a key technology for the Artificial Intelligence of Things (AIoT) to perceive the physical world environment where a reasonable task assignment method can effectively improve the execution efficiency of mobile crowdsensing tasks. Existing task allocation methods cause long delays when dealing with large-scale task allocation, which are inapplicable to time-sensitive tasks. Moreover, existing time-sensitive task assignment methods overlook the attributes of workers, resulting in limited task completion rate. Therefore, an edge-cloud cooperative time-sensitive task assignment method towards AIoT is proposed. Specifically, an edge-cloud collaborate task allocation architecture is designed where task allocation is modeled as an optimization problem with time, space, and cost constraints to break through the traditional cloud center task allocation model. Then a spatiotemporal preference attribute perception method is designed to sense workers’ spatiotemporal preference attributes for increasing the probability of task completion. Furthermore, a benefit-maximized time-sensitive task allocation method is presented. Validation results demonstrate the effectiveness of the proposed method in terms of platform benefit and task completion rate.