Can generative AI help identify peer firms?

• 成果类型: Article; Early Access
• 署名作者: Cao, Yi; Chen, Long; Tucker, Jennifer Wu; Wan, Chi
• 署名单位: George Mason University; State University System of Florida; University of Florida; University of Massachusetts System; University of Massachusetts Boston
• 刊物名称: REVIEW OF ACCOUNTING STUDIES
• ISSN/ISSBN: 1380-6653
• DOI: 10.1007/s11142-025-09892-6
• 发表日期: 2025

摘要

We evaluate how well generative AI can perform an important task-identifying product market competitors (peers). We find that machine-generated peers have a high overlap with the peers identified by human experts as well as with the peers identified by established peer identification systems. Machine-generated peers have high correlations with the focal firm in stock returns, sales growth, and gross profit margin in the subsequent year. The correlations are stronger than those derived from identifying peers by analyzing the similarity of business descriptions in annual reports or by using members in the focal firm's SIC industry. Machine-generated peers also exhibit higher homogeneity among themselves than those identified via the two alternative systems. We demonstrate the usefulness of machine-generated peers in two settings: (1) compensation benchmarking by investors and (2) hypothesis testing by researchers. Overall, our findings suggest that generative AI can identify peer firms reasonably well, especially for large firms.