Damage-involved structural pounding during earthquakes has been recently intensively studied using different impact force models. The results of the previous studies indicate that the linear viscoelastic model is relatively simple yet accurate in modelling pounding-involved behaviour of structures during earthquakes. The only shortcoming of the model is a negative value of the pounding force occurring just before separation, which does not have any physical explanation. The aim of the present paper is to verify the effectiveness of the modified linear viscoelastic model, in which damping term (related to modelling of damage effects) is activated only during the approach period of collision therefore overcoming this disadvantage. The accuracy of the model is checked in a number of comparative analyses, including the comparison with the results of impact experiments and shaking table experiments on pounding between two steel towers. The results of the study indicate that the use of the modified linear viscoelastic model leads to very similar pounding-involved responses as in the case of the linear viscoelastic model.