Oral practice is challenging for foreign language education, and Corrective Feedback (CF) is often used to point out learners' pronunciation errors and to help them improve their oral skills in foreign language courses. CF is generally considered as a necessary condition for foreign language acquisition, and "reflection" and "correction" are imperative stages to realize the process from "input" to "output" to improve accuracy and deepen reflection on foreign language learning. In conventional courses, teachers have often used peer assessment (PA) to ask students to comment on each other's performance during exercises so as to promote reflection. However, due to students' varied levels of proficiency, correction from the teacher during PA activities is needed. Since a ratio of one teacher to many students is very common in most courses, it is almost impossible to provide immediate and detailed feedback for individual students during pronunciation practice. With the development of artificial intelligence, Automated Corrective Feedback (ACF) can provide more personalized, precise, and timely feedback for individual learners. Therefore, a quasi-experimental research design was conducted to explore if the ACF-based peer assessment (ACF-PA) approach would have a positive effect in a foreign language pronunciation course. The participants were 66 college students who were divided into an experimental group with the ACF-PA approach (N = 31) and a control group with the Conventional Peer Assessment (C-PA) approach (N = 35). The results indicated that when ACF-PA was adopted, it was helpful for improving students' learning performance, intrinsic motivation, and self-regulated learning conceptions when learning French pronunciation. Additionally, discussion on students' learning experience and their perceptions is also provided.