A comprehensive evidence map reviewing 73 AI studies in radiopharmaceutical therapy found strong technical performance across segmentation, dosimetry, and prediction tasks. However, it reported no prospective evidence demonstrating that AI systems change treatment decisions in routine clinical workflows. The review highlights a persistent translational gap: algorithms are improving, but rigorous prospective clinical validation remains scarce. That gap limits confidence that technical accuracy will translate into better patient outcomes or altered care pathways. For developers, the findings signal the need for prospective study designs, external validation strategies, and endpoints tied to decision-making rather than imaging metrics alone.