Iron loss is usually neglected in induction motor models used for speed sensorless control and observer design. Thus, the complexity of control and estimator algorithms are reduced. However, the application of a more accurate model can improve the performance, hence the inclusion of iron loss in the machine model is becoming more widespread. But the effect of iron loss modeling on observability has not been analyzed yet. In this paper, observability conditions are presented for nonlinear state-space models with included iron loss. Furthermore, the results are compared with the observability properties of the traditional models where iron loss is ignored.