This paper presents a prototype tool for the space planning phase, which automatically generates alternative floor plans, according to the architect's preferences and desires, and assesses their thermal performance by coupling it with dynamic simulation. A case study of a single-family house was carried out, which comprehended two design sets. The first set correspond to a single-level house and the second set is a two level house served by one stair. Each set is made up of twelve alternative floor plans that were automatically generated, assessed, and ranked according to their thermal performance. The ranking function weights and factors variability is analyzed. The results demonstrate that two level design solutions have the best thermal performance and, within each set, the difference between the best and the worst thermal performance individual may reach 17% in the first set and 35% in the second set. (C) 2014 Elsevier B.V. All rights reserved.