The opportunity to teach students critical thinking is often severely underrepresented in traditional digital forensics curricula. Such an oversight can leave an investigative workforce unable to review digital artefacts with the analytical rigour required in complex, ambiguous cases—where investigators must judge not only what evidence appears to show, but also whether it is trustworthy. As AI-generation technologies (including deepfake generators) become increasingly sophisticated and accessible, the ability of a digital forensics investigator to discern inconsistencies and assess source credibility becomes critical. This paper presents a multi-AI-powered NPC (non-playable character) extension to an existing digital forensics game environment, designed to support educator-authored scenarios featuring multiple witnesses, competing accounts, and evidence-gated disclosures. Drawing on research on epistemic cognition, the extension enables educators to stage multiple sources with distinct evidence, claims, and perspective-bounded disclosures for students to compare and corroborate, rather than treating dialogue as a single authoritative channel. We report the architecture and runtime authoring mechanisms of the extension, and outline a preliminary scholarship of teaching and learning (SoTL)-oriented deployment and descriptive analysis frame grounded in built-in chat, progression, and audit instrumentation. Early observations of the enhanced game, deployed in a postgraduate digital forensics module setting, suggest that the extension gives educators a practical way to target source evaluation and evidence corroboration under uncertainty.