Our paper tackles the development of media reporting during the COVID-19 pandemic, focusing on the January - November 2020 time span, in France, Germany, Romania, Spain, Switzerland, and the United Kingdom. We aim to make media reporting transparent on two dimensions: the coverage of COVID-19-related topics and the negativity of the COVID-19 media reporting. To achieve this goal, we analysed a large news dataset with 841,415 pieces of news—including 202,608 COVID-19 media reports—on an LSTM neural network. The news sentiment data and the corresponding coverage are set in relation to the WHO data on COVID-19 and to Google Trends. This compares the reality, that is WHO data, the perceived and reported reality, that is news data, and the actions based on the perceived and the actual reality, that is Google Trends. The results show that media reporting on COVID-19 is unprecedented in terms of coverage and negativity. Furthermore, the study quantifies how far media reporting detached from the facts after the first wave of COVID-19 and how an Infodemic spread across Europe.