In recent years, there have been substantial efforts from researchers and practioners to improve the trip generatioin model in order to address some of the pitfalls of conventional approaches. These new trip generation models often adopt sophisticated non-linear model forms to utilize new information and incorporate new factors influencing trip generation. However, if sufficient caution is not taken in their application, these new trip generation models may generate severe biases. Two typical sources of biases in the applications of such models are aggregation bias and detransformation bias. The former arises when aggregate variables are included in an individual-level model or when naive aggregation of independent variables are conducted in prediction from a non-linear model, and the latter when de-transforming a non-linearly transformed dependent variable in the prediction process (e.g. predicting from a semi-log model). While these two types of biases are well known and corrections routinely suggested in other disciplines (Wooldridge, 2012, pp212-215) or even in other models of travel demand (Ortuzar and Willumsen, 2011), they have not been adopted in the new trip generation models. In this paper, we describe and demonstrate these two sources of biases with numeric simulations and empirical analyses, before discuss their implications and remedies in applications in the context of modeling trip generation.