Abstract Occupant behavior has an important impact on building energy use while it was seldom taken into consideration in existing building energy benchmarking studies. This may result in an inaccurate benchmarking value and cause a poor building energy performance evaluation. To overcome the limitation, this paper proposes a novel methodology based on the Agent-based modeling method that was able to mimic real occupant behavior in buildings. To demonstrate its applicability, an Agent-based model was developed for a typical household, in which occupant behavior associated with different appliances has been considered. The simulated energy use was utilized as the benchmarking value to further evaluate the building’s energy-saving potential. The results show that, compared with existing methods, the proposed methodology can provide a more practical and reliable building energy benchmarking value, thereby improving the accuracy of identified energy-saving potential. It is suggested that future research should put more effort into developing a more systematic method with different factors being considered such as socioeconomic factors.