This classroom-based study examines whether the use of generative AI for self-editing, combined with explicit instruction, influences CEFR A1–B1 Japanese EFL learners’ writing development and affective factors related to English learning. Participants were 50 first-year students enrolled in two compulsory English courses at a national university in Japan during the second semester of the 2024 academic year. One class was assigned to an experimental group, in which students were required to use ChatGPT for self-editing and received explicit instruction on its effective use, while the control group was allowed optional AI use without explicit guidance. Writing development was measured using pre- and post-test essays assessed by word count and a six-category analytic rubric, and affective factors were examined through questionnaires. A significant increase in word count was observed over time in both groups; however, no significant differences were found between the groups in word count/rubric scores. Increased motivation and self-efficacy were reported in both groups. These findings suggest that over a single semester, writing development among lower-proficiency learners may be influenced more by increased opportunities for learner output than by individualized AI-generated feedback alone. The study discusses implications for CALL-oriented instructional design that leverages AI to promote productive output.