Antimicrobial peptides (AMPs) are recognized for their potential application as new generation antibiotics, however, up to date, they have not been widely commercialized as expected. Although current bioinformatics tools can predict antimicrobial activity based on only amino acid sequences with astounding accuracy, peptide selectivity and potency are not foreseeable. This, in turn, creates a bottleneck not only in the discovery and isolation of promising candidates but, most importantly, in the design and development of novel synthetic peptides. In this paper, we discuss the challenges faced when trying to predict peptide selectivity and potency, based on peptide sequence, structure and relevant biophysical properties such as length, net charge and hydrophobicity. Here, pore-forming alpha-helical antimicrobial peptides family isolated from anurans was used as the case study. Our findings revealed no congruent relationship between the predicted peptide properties and reported microbial assay data, such as minimum inhibitory concentrations against microorganisms and hemolysis. In many instances, the peptides with the best physicochemical properties performed poorly against microbial strains. In some cases, the predicted properties were so similar that differences in activity amongst peptides of the same family could not be projected. Our general conclusion is that antimicrobial peptides of interest must be carefully examined since there is no universal strategy for accurately predicting their behavior.