This study explored how first-year university students used ChatGPT during a semester-long academic writing course on argumentative writing. One hundred and twenty-four students enrolled in a required freshman writing course at a Korean university received explicit instruction in Toulmin-based argumentation and wrote argumentative essays. ChatGPT use for writing was measured across six subscales (surface revision, content elaboration, organization/coherence, metacognitive use, dependence, and collaborative refinement). Analytic rubric scores showed modest overall gains, with particularly strong improvements in evidence and counterarguments. The findings highlight that the quality, rather than mere presence, of AI use matters in strengthening the quality of argumentative writing.