I treat the tendency to misread AI’s deviations and “hallucinations” as truths or signs as a cultural mechanism, rather than simply a technical mistake. Through repeated media translation and an archival approach (AI-generate, select, recompose, re-input), I turn these processes into visual evidence that can be examined and discussed. The projects take a critical position: while creating immersion and attraction, they also reveal how system instability, the agreeable tone of AI language, and the viewer’s own projections can together produce the illusion of a “digital oracle”.