Despite the recent development of automated writing evaluation (AWE) technology and the growing interest in applying this technology to language classrooms, few studies have looked at the effects of using AWE on reducing grammatical errors in L2 writing. This study identified the primary English grammatical error types made by 66 Taiwanese university students in order to examine how using AWE in a process-writing approach affected improvements in these error types in revisions and new text composition. The study adopted a time-series research design, in which measurements of grammatical errors in student writing were taken at various points during a nine-week pedagogical programme. An AWE system was employed as the writing and assessment tool. The AWE feedback reports on the student essays were analysed using descriptive analysis and paired-samples t tests. Using AWE in a process-writing approach seemed to exert a reducing effect on the number of grammatical errors during both revisions and new text composition, although the patterns regarding the precise appearances of effects varied amongst the error types.