Possibility to predict short peptide sequences capable to penetrate the plasma membrane opens new opportunities for developing peptide based intracellular delivery vectors, called cell-penetrating peptides (CPPs). Predictions of CPPs, however are often based on trial and error and may not always lead to new potent sequences. In this review we discuss different problems associated with CPP prediction. Additionally, the used methods of CPP prediction are compared. Also, a few suggestions are made for designing new CPP sequences and improvement of predictions.