As a typical flexible load resource, air conditioning (AC) loads can be aggregated by load aggregator to participate in demand response (DR) markets, effectively alleviating electricity supply pressure during summer peak periods. A hierarchical game-based optimal dispatch strategy for AC aggregators considering response characteristic-based grouping is proposed. First, the characteristics of AC devices are analyzed, and a relationship equation is derived to capture the correlation between average power consumption and temperature difference over long time intervals, thereby revealing the intrinsic relationship between minimum average power and the temperature difference range. Then, considering the heterogeneity of unit-level parameters, an affinity propagation (AP) clustering algorithm is employed to perform fine-grained grouping of AC loads. On this basis, a game-theoretic optimization model is established for AC aggregators. To address solution tractability, continuous intermediate variables are introduced during the solving process, and hierarchical optimization is adopted to ensure the uniqueness of the equilibrium solution. The simulation results demonstrate that the proposed air-conditioning operation equation effectively captures the deep relationship between the minimum hourly operating power of individual units and the temperature-difference interval, thereby reducing the dimensionality of decision variables. Based on the clustering results, the proposed game-theoretic hierarchical optimization strategy exploits the response potential of air-conditioning loads, reducing the peak-to-valley difference of system operation within the scheduling period to 27.22%.