This paper presents an ongoing research project that aims to propose a geopolitical analysis of anti-immigrant speech published on the Mexican twitosphere. While Mexico has long defined itself as an emigration country, the apparent growing presence of anti-immigrant discourse online, especially at the Mexican borders, invites us to question the impact of Americans’ anti-immigrant speech on Mexicans’ representations. We thus propose a transdisciplinary approach that combines a Convolutional Neuronal Network to detect anti-immigrant speech in geolocalized tweets in Mexican Spanish and a geopolitical diachronic analysis to estimate the relationship between such speeches and Americans’ anti-immigrant online representations. With an overall accuracy of 0.76, we are confident that with some improvements the CNN model will be able to detect Mexican anti-immigrant speech on Twitter. We finally discuss that the scope of the analysis would be greatly improved if paired with network and territorial analysis of Mexican anti-immigrant tweets.