“The Immersion Paradox of AI Authorship”: Unveiling the Artificiality Threshold in Non-Core Game Narrative Texts

As generative AI becomes embedded in narrative game production, a central question concerns whether players can sustain emotional immersion once machine authorship is disclosed. This study proposes the Game Artificiality Threshold (GAT), defined as a cognitive–affective boundary where awareness of AI authorship disrupts immersion without reducing perceived textual quality. Grounded in the Computers Are Social Actors (CASA) framework, a two-phase mixed-methods design was employed. Phase I examined player responses to AI-generated non-core texts under varying disclosure conditions, while Phase II introduced a human-authored baseline and expert-rated quality controls. Results from 194 participants indicate that authorship disclosure significantly decreased narrative immersion (p < .01). Symbolic Encoding Consistency (SEC) strongly predicted perceived text quality (β = 0.75, p < .001) but did not mediate disclosure effects. Findings support GAT as a key mechanism in player reception and extend CASA to static narrative contexts, offering implications for transparent yet engaging AI-assisted storytelling.i.