This quasi-experimental study investigated the effect of using model texts as a written corrective feedback (WCF) strategy on second language learners' writing, in both rewritten drafts and new drafts. It further examined learners' perceptions of model-based feedback and how those perceptions affected model effectiveness. Sixty-six English learners were equally divided into either a model group (i.e. an experiment group) or a control group. They wrote an argumentative essay, rewrote the same text, and then wrote a new text using a different prompt. Before rewriting their first draft, the model group received model texts with which they compared their initial writing, whereas the control group self-corrected their errors instead. The model group also completed a perception questionnaire on the usefulness of model-based feedback. The model group outperformed the control group, but only in the rewriting task. Furthermore, in the rewriting task there was a significant correlation between learners' perceptions of models and the effectiveness of model-based feedback, but not in the new writing task.