Presented in this paper is a model-based method for accurately estimating the percent ethanol content of an ethanol-gasoline fuel blend used in flexible fuel vehicles (FFVs). Few methods exist in literature addressing the problem of ethanol estimation in production FFVs. The proposed approach has a distinct advantage as it uses the existing sensor set on a production vehicle. Measurements from the sensors are used to adapt a steady state parametric model relating the engine speed, throttle angle opening and the air-fuel ratio to the fuel injector pulse width command. A first-principles physics-based approach is used to develop the model structure. The fuel composition information is embedded in the adapted model coefficients whose variation is used for ethanol content estimation. The method is validated with experiments on a 2005 Ford 5.4-L V8 port fuel injected engine.