The Impact of Detected AI Reviews on Market Performance: Evidence from Video Game Sales

Abstract

The advent of generative AI technologies is fundamentally transforming content on various online platforms. While recent research has begun to examine their effects on knowledge-sharing behaviors, their economic impact on market outcomes remains poorly understood. We provide initial evidence of this, examining recent customer reviews on Steam.com. Leveraging a comprehensive dataset of 107,198 reviews for the 50 most popular games on Steam.com from June 2021 to November 2023—a period encompassing the launch of ChatGPT—we employ a multi-method approach. Using a novel detection methodology, we find that a higher proportion of AI-generated reviews is associated with a statistically significant increase in game sales. This result is robust to both Ordinary Least Squares (OLS) regression and a Difference-in-Differences (DiD) design that exploits the exogenous shock of ChatGPT's release to establish causality. Furthermore, we find that this positive relationship is significantly moderated by game genre, suggesting that the effectiveness of AI-generated content is context-dependent. These findings have important implications for game developers and platform owners, helping inform decisions about whether to adopt a welcoming or preventive approach to generative AI use based on specific product characteristics.

Keywords

Generative artificial intelligence, online review, detected AI reviews