Common sense tells that historical data are more informative for the estimation of today’s nowcasting models when observed in a similar economic state as today. We operationalise this intuition by proposing a state-based weighted estimation procedure of GDP nowcasting models, in which observations are weighted based on the similarity with the current economic state. For this end, we also construct new state variables for measuring the similarity of economic time periods using news data, in addition to traditional variables. We find that the state-based weighting of observations leads to economically significant nowcasting gains for predicting GDP growth in Belgium as compared to traditional unweighted approaches.