Cancer therapy is currently limited by the difficulty of achieving efficient delivery into target cells. To investigate whether therapeutics can be delivered specifically to cancer cells, we have explored the possibility of selecting small peptides that bind specifically, or preferentially, to breast cancer cell lines. By using random peptide phage libraries and an experimental approach that allows the selection of internalized peptides, cell-specific binding peptides have been identified. The peptides define a major core motif (LTVXPWY) that was not found in negative phages. Phage displaying LTVSPWY peptide sequence exhibited a specific binding to breast cancer cells. None of the selected peptides bound to human primary cells from different tissue origin (e.g., epithelial, endothelial, hematopoetic). The potential of the selected peptides to mediate cellular internalization in the context of phages and recombinant GFP-peptide fusions was demonstrated. By linking the LTVSPWY peptide to an antisense phosphorothioate oligonucleotide against the ErbB2 receptor, specific delivery to cancer cells was achieved. In contrast to free antisense, the peptide-antisense conjugates inhibited ErbB2 gene expression. Thus, efficient delivery of antisense oligonucleotides can be achieved by coupling them to cancer cell-specific peptides, identified by a method that did not require any knowledge about their corresponding receptors.