‘Digital twins’ are widely invoked yet often loosely defined, carrying a range of techno-optimistic promises. The term spans sectors—from smart cities and aerospace to healthcare, logistics, and software engineering—but twinning also operates as a broader cultural and technical logic, appearing in dashboards, synthetic data, generative AI, deepfakes, and personalized systems such as AI agents. Said to mirror reality in order to optimize it, twinning promises accurate reflection of systems, environments, and even people. But what if the twin is read not just as a doppelgänger, but as a complex relational entity that rearranges sameness and difference? In an age intent on rendering the world fully computable, the notion of twinning is attractive because it suggests fully knowable, manageable, and optimizable systems: a world fully represented, a person fully captured. This keynote invites a shift in perspective—from complete likeness to absence and difference, from doppelgänger to persona—and a critical reflection on the consequences that ‘false’ twins introduce. Consider non-consensual deepfakes often labelled ‘Not Taylor Swift.’ These pieces of synthetic content stage personas calibrated to be just inaccurate enough to linger in legal and ethical grey zones. In such cases the twin becomes an alibi: similar enough to perform, different enough to deny accountability. Drawing on examples of digital twins across diverse domains, Randi Heinrichs’s talk shows that twinning is not simply representation; it is a logic that plays with recognition and misrecognition. Digital twins actively compose worlds, objects, and subjects: they classify, extract, approximate, and regenerate. As Heinrichs illustrates, such entities do not merely reflect but perform, intervene, and reshape the realities they capture, with profound political and ethical consequences.