In recent years, energy-related CO2 emissions embodied in international trade and the driving forces have been widely studied by researchers using the environmental input–output framework. Most previous studies however, do not differentiate different input structures in manufacturing processing exports and normal exports. Using China as an example, this paper exemplifies how implications of results obtained using different export assumptions differ. The study posits that the utilization of traditional I–O model results in an overestimation of emissions embodied in processing exports and an underestimation in normal exports. The estimate of CO2 emissions embodied in China's exports drops by 32% when the extended I–O model is used. The choice of export assumption has more impact on the decomposition results for processing exports. The study further highlights that for a country with an export structure similar to China, it is meaningful to look into the impact of export assumption in embodied emission studies. ► We study the issue of exports assumptions in embodied emission studies and structural decomposition analysis. ► The implications of the results obtained using two different exports assumptions are not the same. ► Utilization of traditional I–O model results in an overestimation of emissions embodied in processing exports. ► Utilization of traditional I–O model results in an underestimation of emissions embodied in normal exports. ► The choice of exports assumption has more impact on the decomposition results for processing exports.