Genetic materials are exchanged very frequently across microbial organisms. In many cases, a cluster of adjacent genes, rather than individual genes, are transferred from one genome into another genome. The influx gene clusters in the recipient genome are also known as genomic islands (GIs). While various computational approaches have been employed to identify genomic islands, they are not universally applicable to predict any microbial genome. In this paper, we propose a genetic algorithm (GA) approach to search genomic islands for any genome. The GA evaluates candidate solutions in the population based on sequence composition based fitness value, and finds final solutions through crossover and mutation operations throughout the evolution. We applied our GA approach on four genomes. Experimental results have shown that our GA could not only find previously reported GIs, but also find some new GIs not reported before. We believe that our GA-based approach for GI finding should complement existing GI prediction tools.