Despite widespread cervical cancer (CC) screening programs, low participation has led to high morbidity and mortality rates, especially in developing countries. Because early-stage CC often has no symptoms, a non-invasive and convenient diagnostic method is needed to improve disease detection. In this study, we developed a new approach for differentiating both CC and cervical intraepithelial neoplasia (CIN)2/3, a precancerous lesion, from healthy individuals by exploring CC fatty acid metabolic reprogramming. Analysis of public datasets suggested that various fatty acid metabolizing enzymes were expressed at higher levels in CC tissues than in normal tissues. Correspondingly, 11 free fatty acids (FFAs) showed significantly different serum levels in CC patient samples compared with healthy donor samples. Nine of these 11 FFAs also displayed significant alterations in CIN2/3 patients. We then generated diagnostic models using combinations of these FFAs, with the optimal model including stearic and dihomo-γ-linolenic acids. Receiver operating characteristic curve analyses suggested that this diagnostic model could detect CC and CIN2/3 more accurately than using serum squamous cell carcinoma antigen level. In addition, the diagnostic model using FFAs was able to detect patients regardless of clinical stage or histological type. Overall, the serum FFA diagnostic model developed in this study could be a powerful new tool for the non-invasive early detection of CC and CIN2/3.