Protein switches have a variety of potential applications in biotechnology and medicine that motivate efforts to accelerate their development. Switches can be built by the proper fusion of two proteins with the prerequisite input and output functions. However, the exact fusion geometry for switch creation, which typically involves insertion of one protein domain into the other, is difficult to predict. Based on our previous work developing protein switches using periplasmic binding proteins as input domains, we wondered whether there are "hot spots" for insertion of output domains and successful switch creation within this class of proteins. Here we describe directed evolution experiments that identified switches in which TEM-1 beta-lactamase (BLA) is inserted into the class I periplasmic binding proteins ribose binding protein (RBP), glucose binding protein (GBP), and xylose binding protein (XBP). Although some overlap in sites for successful switch insertion could be found among the paralogs, successful switches at these sites required different linkers between the domains and different circular permutations of the BLA protein. Our data suggests that sites for successful switch creation are not easily transferable between paralogs. Furthermore, by comparison to a previous study in which switches were created by inserting a xylanase into XBP, we find no correlation between insertion sites when using two different output domains. We conclude that the switch property likely depends on the precise molecular details of the fusions and cannot be easily predicted from some overall general structural property of the fusion topology.