Cancer-associated venous thrombosis (CAT) is a common, multifactor event known to complicate the course of cancer and jeopardize a patient's prognosis. The current guidelines regarding the prevention of CAT are sometimes considered insufficiently precise about specific situations, or are poorly applied. The expected benefits of thromboprophylaxis are balanced by the risk of major bleeding induced by anticoagulation, which implies a need to accurately identify ambulatory patients at high risk of thrombosis or hemorrhage. The Khorana score is commonly used for this, but is limited by the non-reproducibility of predicted performance across cancer types, and by the fact that antitumor treatment and cardiovascular risks are not included. The COMPASS-CAT score, which includes those two aspects, was found to be a more accurate predictor of venous thromboembolism in patients with lung cancer, and to better distinguish between patients at low or high risk of thrombosis. The frailty of patients with cancer is also a major issue, and should be taken into account when thromboprophylaxis is considered. According to current guidelines, CAT prophylaxis should be considered for hospitalized patients, those for whom surgery is scheduled, or those with pancreatic cancers. In ambulatory patients, decisions should be made according to patient, cancer and antitumoral treatment characteristics. Low molecular weight heparin is the gold standard of CAT prophylaxis. Despite increased risks of bleeding or drug-drug interactions in cancer patients, direct oral anticoagulants could be alternate options for high-risk ambulatory patients that should be accompanied by a careful global analysis of benefits, harms, and patient preferences.