Purpose To investigate the usefulness of the positron emission tomography response criteria in solid tumors 1.0 (PERCIST1.0) for predicting tumor response to neoadjuvant chemotherapy and prognosis and determine whether PERCIST improvements are necessary for esophageal squamous cell carcinoma (ESCC) patients. Patients and methods We analyzed the cases of 177 ESCC patients and examined the association between PERCIST and their pathological responses. Associations of whole-PERCIST with progression-free survival (PFS) and overall survival (OS) were evaluated by a Kaplan-Meier analysis and Cox proportional hazards model. To investigate potential PERCIST improvements, we used the survival tree technique to understand patients’ prognoses. Results There were significant correlations between the pathologic response and PERCIST of primary tumor ( p  < 0.001). The optimal cutoff value of the primary tumors’ SULpeak response to classify pathologic responses was −50.0%. The diagnostic accuracy of SULpeak response was 87.3% sensitivity, 54.1% specificity, 68.9% accuracy, positive predictive value 60.5%, and negative predictive value 84.1%. Whole-PERCIST was significantly associated with PFS and OS. The survival tree results indicated that a high reduction of the whole SULpeak response was significantly correlated with the patients’ prognoses. The cutoff values for the separation of prognoses were − 52.5 for PFS and − 47.1% for OS. Conclusion PERCIST1.0 can help predict tumor responses and prognoses. However, 18 F-FDG-PET/CT tends to underestimate residual tumors in histopathological response evaluations. Modified PERCIST, in which the partial metabolic response is further classified by the SULpeak response (−50%), might be more appropriate than PERCIST1.0 for evaluating tumor responses and stratifying high-risk patients for recurrence and poor prognosis.