Metastatic carcinoma of unknown primary location contributes to approximately 4% of all diagnosed carcinomas. Some metastatic carcinomas have distinctive histologic features that allow for their site determination (e.g., colonic adenocarcinoma, bronchioloalveolar cell carcinoma), though the often the features that are not distinctive enough to determine the site of origin. In addition, a histopathological distinction of primary and secondary tumors may also complicate the diagnosis. Pathologists are now frequently asked to determine the primary site for metastatic carcinomas of unknown origin using adjunctive techniques, such as immunohistochemistry and even molecular alterations. The importance of judicious attention to subtle morphological features usually provides highly valuable clues to the site of the tumor, which may substantially narrow the diagnostic possibilities. The purpose of this article is to present an algorithmic approach to morphologic classification of carcinomas of unknown primary site. Emphasis is placed on the cytomorphologic features of the malignancy, with consideration to the patient’s age, anatomic distribution of disease, and histochemical studies.