Purpose: Generation of antibodies which potentially discriminate between malignant and healthy cells is an important prerequisite for early diagnosis and treatment of gastric cancer (GC). Comparative analysis of cell surface protein landscape will provide an experimental basis for biomarker discovery, which is essential for targeted molecular therapies. This study aimed to isolate phage-displayed antibody fragments recognizing cell surface proteins, which were differently expressed between two closely related GC cell lines, namely AGS and MKN-45. Methods: We selected and screened a semisynthetic phage-scFv library on AGS, MKN-45, and NIH-3T3 cell lines by utilizing a tailored selection scheme that was designed to isolate phagescFvs that not only recognize the differentiated AGS cells but also distinguish them from NIH3T3 fibroblasts and the poorly differentiated MKN-45 cells. Results: After four rounds of subtractive whole cell panning, 14 unique clones were identified by ELISA screening and nucleotide sequencing. For further characterization, we focused on four phage-scFvs with strong signals in screening, and their specificity was confirmed by cell-based ELISA. Furthermore, the selected phage-scFvs were able to specifically stain AGS cells with 38.74% (H1), 11.04% (D11), 76.93% (G11), and 69.03% (D1) in flow cytometry analysis which supported the ability of these phage scFvs in distinguishing AGS from MKN-45 and NIH-3T3 cells. Conclusion: Combined with other proteomic techniques, these phage-scFvs can be applied to membrane proteome analysis and, subsequently, identification of novel tumor-related antigens mediating proliferation and differentiation of cells. Furthermore, such antibody fragments can be exploited for diagnostic purposes as well as targeted drug delivery of GC.