Performance feedback is one of the most critical ways to motivate and facilitate individual progress. With the advancement of artificial intelligence (AI) technology, feedback provided by AI is increasingly applied in practice, gradually surpassing the quality of performance feedback delivered by human managers. It has become a significant topic in organizational management research. However, existing literature on AI feedback is scattered across different disciplines such as organizational management, education, and healthcare. This dispersion has led to significant differences in research paradigms, theoretical perspectives, and empirical methods within the study of AI feedback. Furthermore, existing literature has not yet formed a unified understanding of the theoretical mechanisms behind the varying effects of AI feedback. In light of this, this paper first clarifies the concept of AI feedback. Next, it systematically summarizes and reviews the deployment effect and disclosure effect through which AI feedback exerts its influence, thereby constructing a research framework for AI feedback. Then, it introduces and summarizes the commonly used or profoundly insightful theoretical mechanisms in existing AI feedback research and discusses the ways these mechanisms are applied. Finally, it proposes five future research directions that hold both scientific value and practical significance.