AI-Style Expression and Game Feedback Discriminability: Recommendation Dynamics and Review Information Prioritization

AI-style expression has become a feature of online review environments, raising questions about how clearly review content continues to differentiate recommendation states. Using 108,911 English-language Steam reviews from 18 games, this study defines Game Feedback Discriminability (GFD) as the out-of-sample separability of recommendation labels from masked review content. In competitive-action games, larger increases in AI-style expression are followed by a higher next-quarter Not Recommended share; the narrative-choice association is near zero, with suggestive support for a direct type difference. Under the actual-calendar specification, the within-type AI-style–GFD relationship follows different temporal paths across the two game types: the narrative-choice relationship moves from negative toward zero, whereas the competitive-action relationship declines from positive to negative. Prioritizing game content and operational information is associated with higher GFD in both types. In competitive-action games, evaluation-linked gameplay experience is associated with higher GFD, whereas problem-focused information is associated with lower GFD.