*with Josep Amer-Mestre and Marta C. Lopes* - **Applied Economics** (2023). DOI:10.1080/00036846.2023.2167924 We study the impact of COVID-19 school closures on differences in online learning usage by regional academic performance. Using data from Google Trends in Italy, we find that during the first lockdown, regions with a previously lower academic performance increased their searches for e-learning tools more than higher-performing regions. Analysing school administrative and survey data before the pandemic, we find that both teachers and students in lower performing regions were using no less e-learning tools than higher performing ones. These two findings suggest that the COVID-19 shock widened the e-learning usage gap between academically lower and higher-performing regions. Exploiting the regional variation in school closure mandates during the 2020/2021 academic year, we report that the patterns detected after the first lockdown were no longer present. Regions with different previous academic performance had the same response in terms of online learning usage when faced with stricter school closures.