Little is known about how monolingual ideologies and their effects manifest in online contexts as compared to offline contexts. We conducted a corpus-assisted discourse study to investigate this, with a focus on Twitter representations of Spanish as a heritage language in the USA. We analysed two corpora (one English and one Spanish--over 30 million words), examining frequencies, collocations, concordance lines, and larger text segments. The results revealed evidence of the same ideologies found in offline contexts: normative monolingualism (drawing on the one-nation-one-language ideology and language purity ideologies) as well as raciolinguistic ideologies. We show how the semiotic processes of iconization and erasure lead to the (evidently erroneous) essentialization of Spanish heritage language speakers as a homogeneous group of un-American, racialized immigrants with broken language. This discursive creation of difference constitutes the basis for the systematic discrimination of Spanish heritage language speakers, thus reflecting and reproducing social inequality. Our findings therefore indicate the necessity of extending planning measures to protect heritage language speakers (and other minority groups) from offline contexts to online contexts.