The communities that built their identity around exposing AI fakes have become unreliable narrators of their own cause. Public retractions on Bluesky — rare enough to be embarrassing, common enough to form a pattern — signal that the detection skills driving the accusations are not keeping pace with the confidence behind them. This is not a story about bad-faith actors; most of the people walking back their threads believed what they said when they said it. The problem is structural: the epistemology of AI suspicion rewards speed over accuracy, and…