Many languages exhibit differential object marking (DOM), where only certain types of objects are marked. Traditionally, it has been claimed that DOM arises as a way to prevent ambiguity by marking objects that might otherwise be mistaken for subjects (e.g., animate objects). However, a recent account proposes that DOM may instead arise as a way to mark objects that are atypical in terms of information structure. According to this account, rather than being marked to avoid ambiguity, objects are marked when they are given (already familiar in the discourse) rather than new. Recent experimental work supports this view (Tal et al., 2022. Importantly, it has been argued that even semantically-conditioned DOM systems (e.g., in which animate objects receive marking) originally stem from atypical information structure (Cristofaro, 2013, Iemmolo, 2010). According to this account, since given objects tend to also be animate, the givenness-conditioned marking is reanalyzed over time as animacy-conditioned. We conducted an artificial language learning experiment to test this hypothesis. While we replicated previous findings showing that atypical information structure can lead to DOM systems, we did not find evidence for grammaticalization to animacy-conditioned marking. We discuss possible reasons for these findings and their implications for our understanding of DOM systems and their emergence.