We propose a spatio-temporal analysis aiming at estimating the variation of gasoline and diesel consumption occurred in the Italian provinces as a consequence of the COVID-19 pandemic during 2020 and 2021. We employ generalized linear models and generalized additive models to predict business-as-usual per capita fuel consumption by means of socio-economic variables, such as demography, tourist stays, temperature and territorial characteristics. The province-specific impact of the restrictions is then estimated as the difference between the observed and model-predicted values. The spatio-temporal dynamics is introduced by means of appropriate spline basis functions allowing both spatial and time trend to be smooth surfaces. Moreover, we include a smooth spatio-temporal interaction that properly capture the seasonal characteristic of different areas. Our findings highlight that the national mobility restrictions imposed to fight the spread of COVID-19 in the first wave (March to May 2020) reduced gasoline and diesel consumption homogeneously nationwide, with some slight differences between northern and southern provinces. However, the second wave lockdown (Winter 2020–2021) determined heterogeneous impacts across Italian provinces. In particular, we highlight the existence of a divergence between North and South Italy in terms of reaction to the pandemic. Indeed, while southern provinces experienced the weakest and less variable reductions, the northern provinces recorded strong and persistent contractions of fuel consumption. Also, southern and tourism-devoted provinces recovered quickly and compact to the post-pandemic reopenings.