Eliminating poverty is a multifaceted global challenge and a focal point of global development governance. With the implementation of targeted poverty alleviation (TPA), China's poverty alleviation efforts have had significant achievements. This study provides a new perspective by classifying poverty-stricken households based on poverty root causes at the household level, a multi-propensity score weighting model based on counterfactual inference is employed to examine the poverty reduction effect and policy precision on six non-equivalent poverty-stricken household groups in a Chinese county. The results reveal that the poverty reduction effects differed among poverty-stricken households with different root causes of poverty. Specifically, households impoverished due to disability have the lowest income of the six groups. The assistance policy effects also vary significantly and are found to be poorly suited to poverty-stricken households a lack of labor force and funds. A robustness test confirms this conclusion and a more nuanced analysis reveals that these differences are reflected in the wage and transfer incomes. Therefore, to ensure the stability of poverty reduction and the sustainability of income for poverty-stricken households, relevant associated aid policies need to place different emphases based on their household characteristics.