Cancer is one of the most fatal diseases threatening public health globally, and tumor metastasis causes greater than 90 % of cancer-associated deaths, presenting huge challenges for detection and efficient treatment of various human cancers. Cancer stem cells (CSCs) are a rare population of cancer cells and increasing evidences indicated CSCs are the driving force of tumor metastasis. In this study, a p-AuNSs-assisted single-cell Raman spectra has been established, to extract and amplify of CSCs fingerprints with single cell sensitivity. The gathered rich molecular information was further assigned according to the characteristic Raman peaks, and the results revealed a huge difference in the expression of molecules between CSCs and cancer cells, such as nucleic acids, proteins, saccharides and lipids. Furthermore, multiple data analysis algorithms including PCA, SVM, RF and KNN, were employed to reveal the fundamental characteristics and classification of CSCs and cancer cells based on the whole p-AuNSs-assisted single-cell Raman spectra. This work is beneficial for not only providing deep insights for molecular behaviors of tumor metastasis based on CSCs, but also could obtain significant information to aid medical diagnosis and to design effective CSCs-based therapeutic intervention clinically.