With diminishing tariffs, the focus of trade policy makers and analysts is logically turning towards non-tariff barriers (NTBs), but there much remains to be done. It is well-known that tackling NTBs poses many challenges for the analyst because of their diverse and complex nature, and the lack of available evidence. NTBs pose also particular difficulties to computable/applied general equilibrium (CGE) modelling, traditionally more comfortable with policies whose impact can be interpreted into direct effects on prices. This research provides a quantification of the impact of NTBs at the global level. The model we use is significantly larger than in previous studies, and to our knowledge, the first truly global one. The main focus of this research, however, is to discuss and question the various treatments of NTBs in CGE models with a specific application of the Global Trade Analysis Project (GTAP) standard model, the workhorse of trade policy CGE analysis. We find that serious estimation and modelling efforts remain to be undertaken in order to make CGE modelling a useful policy tool to analyze NTBs. Casual policy inferences from loose specifications may indeed lead to serious analytical mistakes. We show that while using the same robust estimates of NTB incidence we obtain vastly different results under different model specifications.