Background Although almost every state medical marijuana (MM) law identifies cancer as a qualifying condition, little research supports MM's use in oncology. We hypothesized that the discrepancy between these laws and the scientific evidence base poses clinical challenges for oncologists. Oncologists' beliefs, knowledge, and practices regarding MM were examined in this study. Methods In November 2016, we mailed a survey on MM to a nationally-representative, random sample of 400 medical oncologists. Main outcome measures included whether oncologists reported discussing MM with patients, recommended MM clinically in the past year, or felt sufficiently informed to make such recommendations. The survey also queried oncologists' views on MM's comparative effectiveness for several conditions (including its use as an adjunct to standard pain management strategies) and its risks compared with prescription opioids. Bivariate and multivariate analyses were performed using standard statistical techniques. Results The overall response rate was 63%. Whereas only 30% of oncologists felt sufficiently informed to make recommendations regarding MM, 80% conducted discussions about MM with patients, and 46% recommended MM clinically. Sixty-seven percent viewed it as a helpful adjunct to standard pain management strategies, and 65% thought MM is equally or more effective than standard treatments for anorexia and cachexia. Conclusion Our findings identify a concerning discrepancy between oncologists' self-reported knowledge base and their beliefs and practices regarding MM. Although 70% of oncologists do not feel equipped to make clinical recommendations regarding MM, the vast majority conduct discussions with patients about MM and nearly one-half do, in fact, recommend it clinically. A majority believes MM is useful for certain indications. These findings are clinically important and suggest critical gaps in research, medical education, and policy regarding MM.