<p>With the advent of digitalization, the internet has become a key driver of low-carbon agricultural production. This paper employs survey data from 1192 farmers in China’s three major grain-producing provinces to evaluate agricultural carbon efficiency by applying the Super-SBM model. We applied the Treatment Effect Model (TEM) and the Extended Regression Model (ERM) to examine the effects of internet use on agricultural carbon efficiency, analyzing the underlying mechanism from a factor allocation perspective. The findings demonstrate that internet use significantly enhanced agricultural carbon efficiency by 1.800% compared to farmers who did not use the internet, and each incremental unit of internet usage intensity could improve agricultural carbon efficiency by 2.200%. These improvements are particularly pronounced among farmers with lower levels of agricultural carbon efficiency. The mediating effects of land, labor, and capital allocation in the internet adoption decisions to improve agricultural carbon efficiency were 5.329%, 13.564% and 18.613%, respectively. Similarly, land and capital allocation accounted for 8.582% and 11.835% of the mediating effects of internet usage intensity on agricultural carbon efficiency, while labor allocation had no significant effect. Based on these findings, policymakers are advised to prioritize rural digital infrastructure development and implement targeted programs to promote internet use, particularly among farmers with lower levels of agricultural carbon efficiency. Furthermore, governments should establish transparent land transfer platforms, cultivate a skilled agricultural workforce through vocational training, and promote long-term capital investments in agriculture to bolster low-carbon practice.</p>