Censorship and domain adaptation significantly undermine the effectiveness of automated detection methods in identifying machine-generated tweets.