This paper presents a semi-automatic visualization method for the evaluation of urban environments that is based on artificial intelligence. It proposes the use of agent-based crowd simulation software on a mid-scale urban planning level for design evaluation. The information on agentsi movements is noted in standard raster images. The results are maps that are easy to understand. These maps show movement paths of the agents and density and give further conclusion on bottlenecks in planning contexts. Key measures, like the occupant movement in a given district, until now relied greatly on empirical knowledge or data that could be only gathered after an urban design had become built reality. Our method focuses on the adaptation of common software technology that is originally situated in film and TV productions. A practical workflow shows how our method can be easily integrated in daily design tasks.