The relationships at the base of the metazoan tree have been difficult to robustly resolve, and there are several different hypotheses regarding the interrelationships among sponges, cnidarians, ctenophores, placozoans, and bilaterians, with each hypothesis having different implications for the body plan of the last common ancestor of animals and the paleoecology of the late Precambrian. We have sequenced seven nuclear housekeeping genes from 17 new sponges, bringing the total to 29 species analyzed, including multiple representatives of the Demospongiae, Calcarea, Hexactinellida, and Homoscleromorpha, and analyzed a data set also including six nonmetazoan outgroups and 36 eumetazoans using a variety of phylogenetic methods and evolutionary models. We used leaf stability to identify rogue taxa and investigate their effect on the support of the nodes in our trees, and we identified clades most likely to represent phylogenetic artifacts through the comparison of trees derived using different methods (and models) and through site-stripping analyses. Further, we investigated compositional heterogeneity and tested whether amino acid composition bias affected our results. Finally, we used Bayes factors to compare our results against previously published phylogenies. All our maximum likelihood (ML) and Bayesian analyses find sponges to be paraphyletic, with all analyses finding three extant paraphyletic sponge lineages, Demospongiae plus Hexactinellida, Calcarea, and Homoscleromorpha. All but one of our ML and Bayesian analyses support the monophyly of Eumetazoa (here Cnidaria + Bilateria) and a sister group relationship between Placozoa (here Trichoplax adhaerens) and Eumetazoa. Bayes factors invariably provide decisive support in favor of poriferan paraphyly when compared against either a sister group relationship between Porifera and Cnidaria or with a monophyletic Porifera with respect to a monophyletic Eumetazoa. Although we were able to recover sponge monophyly using our data set, this was only possible under unrealistic evolutionary models, if poorly performing phylogenetic methods were used, or in situations where the potential for the generation of tree reconstruction artifacts was artificially exacerbated. Everything considered, our data set does not provide any support for a monophyletic Diploblastica (here Placozoa + Cnidaria + Porifera) and suggests that a monophyletic Porifera may be better seen as a phylogenetic artifact.