Ranked lists excel at finding candidates but provide no mechanism for judging them. In trademark clearance, where practitioners must assess likelihood of confusion and registrability risk across visual, phonetic, and conceptual dimensions under goods and services constraints, the core task is not retrieval but multi-criteria comparative judgment informed by examination practice and prior prosecution records. We propose agentic simulation as an IR interface and demonstrate it in T-RADAR, a trademark clearance system. Beyond retrieval, T-RADAR introduces simulation-driven adjudication: candidates retrieved via multimodal dual-encoder search can be examined through an adversarial agent-to-agent Examiner versus Applicant exchange governed by a structured protocol of claims and rebuttals, producing a scored assessment that draws on similarity factors described in trademark examination guidelines and serves as a heuristic review prioritization signal. The interaction proceeds in three steps. Query: users input a candidate mark and its goods designation to retrieve potentially conflicting registrations. Select and Simulate: users choose high-risk pairs and launch simulated examination. Refine and Re-simulate: users adjust the mark elements or goods designation based on system suggested strategies and compare how outcomes change.