Machine learning models are widely used for diverse applications, including image classification, natural language analysis, and creating personalized recommendations. To generate personalized recommendations, recommender systems can use supervised or reinforcement learning to predict which items interest a user. At the same time, machine learning models are vulnerable to adversarial attacks. Such attacks attempt to intentionally reduce a model’s performance, for instance, by applying specific changes to an input. Causal machine learning is a promising approach to machine learning that may improve a model’s robustness by learning causal relations instead of correlations. In this paper, the robustness of causal machine learning models against data poisoning attacks has been compared with statistical machine learning models in the context of recommender systems. Based on our results, the causal deep reinforcement learning model CDT4Rec is not more robust against reward poisoning attacks than the evaluated statistical reinforcement learning model DDPG.