Micro-Analysis: Trends in Player Sentiment Toward AI Companions in Modern Games

This study presents first micro-analysis of player sentiment toward AI companions in commercially successful video games. A custom dataset of 15,412 Steam reviews was constructed and used, which contained discourse regarding AI companions. It spanned 17 titles across RPG, JRPG, open-world, shooter, and cinematic action-adventure genres. The research examines how players evaluate companion behavior, personality, and performance. Lexical exploratory data analysis (using unigram, bigram, and keyword co-occurrence analysis) and VADER sentiment analysis were performed to identify sentiment patterns and recurring evaluative themes. Results indicate that functional competence in combat, coordination, and survival is the baseline expectation for positive perception, while mechanical inabilities drive negative sentiment. Emotional integration and narrative embedding were found to amplify positive sentiment and to moderate technical shortcomings. The findings further demonstrate that companion discourse density and assessment are genre sensitive. Systemic companions highlight performance-oriented evaluation, whereas narrative companions are evaluated on the basis of relational depth and character identity. Importantly, companion-specific sentiment emerges independent of overall sentiment for the game. Through empirical evidence gathered, the study extends current literature while validating companion design concepts.