This paper introduces a graphics layout design tool that has a capability to learn user preferences and to recognize designing consistencies. The introduced tool has ability to generate layout templates for selection according to user preferences learned from previous interactions. Additional objects suggestion to preserve consistencies can be made by the system. User preferences are in form of spatial relations and transitive relations. The system employs genetic programming to generate the layouts for selection and to be a learning mechanism. The system architecture, system learning, and experimental results are also included.