Graphophonemic conversion rules are posited by models of word reading in order to handle reading of unknown letter strings such as nonwords. Graphophonemic conversion is thought to be relied upon by readers especially in more transparent orthographies such as Greek. Here we test the hypothesis that Greek nonwords are read using rules, by exploiting a case of orthographic ambiguity found in Greek orthography. Participants read nonwords, some of which resembled their source words while others did not. Rule-based performance predicts uniform treatment of the stimuli. Results showed that readings were heavily influenced both by individual lexical items (when similar) and by distributional properties of the lexicon (majority readings of the ambiguous sequences), interpreted as neighborhood effects, as well as by additional factors. The findings are not consistent with exclusive reliance on graphophonemic conversion rules and point instead to emphasizing the role of the lexicon in providing the substrate for both regular and irregular reading performance.