In previous studies, it has been reported that rosiglitazone has opposing effects on nonalcoholic fatty liver disease. The purpose of the current study is to test the hypothesis that such opposing effects are related to different levels of peroxisome proliferator-activated receptor gamma (PPAR-γ) in the liver. Using a gene transfer approach and mice fed a high-fat diet (HFD) as an animal model, we demonstrate that mice with low levels of PPAR-γ expression in the liver are resistant to HFD-induced development of fatty liver when treated with rosiglitazone. Conversely, rosiglitazone treatment actually exacerbates liver steatosis in obese mice that have a higher level of PPAR-γ. Mechanistic studies show that an elevated hepatic PPAR-γ level is associated with an increased expression of genes responsible for lipid metabolism in the liver, particularly Cd36, Fabp4, and Mgat1. The concurrent transfer of these three genes into the mouse liver fully recapitulates the phenotypic change induced by the overexpression of PPAR-γ. These results provide evidence in support of the importance of PPAR-γ in the liver when rosiglitazone is considered for the treatment of fatty liver disease. Clinically, our results suggest the necessity of verifying PPAR-γ levels in the liver when rosiglitazone is considered as a treatment option, and indicate that the direct use of rosiglitazone for treatment of nonalcoholic fatty liver may not be desirable when the patient’s PPAR-γ level in the liver is significantly elevated.