Pancreatic cancer has a poor prognosis after complete macroscopic resection combined with chemotherapy. Even after neoadjuvant chemotherapy, R0 resection is often not possible. Moreover, current imaging techniques cannot reliably distinguish viable cancer cells from scar tissue at the resectional margin. We investigated the use of a conditionally replicative adenovirus (CRAd), Ad5/3Cox2CRAd-ΔE3ADP-Luc, for imaging the effects of chemotherapy. The CRAd infectivity of pancreatic cancer cells was enhanced by a chimeric Ad5/3 fiber, E1A expression was under the control of the Cox2 promoter, and the luciferase gene was inserted adjacent to the adenovirus death protein (ADP) gene. Subcutaneous xenografts of the pancreatic cancer cell line MiaPaCa-2 were established in 24 BALB/c nu/nu mice. When xenografts reached a diameter of 4–6 mm (day 1), the mice were injected i.p. with either PBS (group A; n = 12) or 1000 mg/kg gemcitabine (group B; n = 12), weekly. On days 19, 26, 33, and 40, CRAd were injected intratumorally into three mice in groups A and B. Bioluminescence was imaged 72 h after CRAd injection, and gross tumor volumes were measured then tumors were removed for ex vivo histopathology using H&E and Ki-67 staining. Correlations between gross tumor volume, pathological evaluation of the percentage of viable tumor area, and CRAd bioluminescence were analyzed. Bioluminescence correlated closely with the percentage of viable tumor area (R = 0.96), but not with gross tumor volume (R = 0.31). Therefore, CRAds might be reliable imaging tools for monitoring chemotherapy in pancreatic cancer, and could improve our ability to distinguish viable tumor cells from scar tissue.