The Italian National Map of Soil Consumption, published by the Italian Institute for Environmental Protection and Research (ISPRA - Istituto superiore per la protezione e la ricerca ambientale), is updated yearly basing on the visual interpretation of soil consumption changes in aerial and/or satellite high resolution images handled by each Regional Environmental Agencies. Various approaches have been tested in the past years by many authors to speed up the visual interpretation process by highlighting areas with the highest probability of land cover changes: in this work a pure deep learning-based method has been tested with Sentinel-2 satellite images, using past 10 m raster maps in the training process. A simple binary classifier with two convolutional layers has been used to create single Sentinel-2 tile raster soil consumption maps using a CPU with and without CUDA support to ensure usability with low-segment workstation in the WinPython portable Python environment. After imposing an appropriate threshold over the classified results obtained for 2022 and 2023 spring Sentinel-2 single images’ shots, the resulting maps has been used in visual interpretation of artificial area changes: while the former has been used in checking 2022 misinterpretations in soil consumption detection maps, the latter has been used to identify high probability areas with soil consumption changes. The resulting probability map, compared with other binary classifiers, while showing in some cases biases in natural areas like the ones founded by using spectral indices, has been found to be a valuable aid for the visual interpretation of high-resolution aerial images in both current soil consumption’ misinterpretation fixing and new soil consumption finding. Results from popular deep learning models used in literature have also been tested during the visual interpretation phase.