Carbon dioxide removal (CDR) is key to climate change mitigation, yet implications of its deployment remain unclear. Recent exponential growth in literature is rapidly filling this gap but makes the synthesis of the evidence on CDR side effects increasingly challenging. We address this issue by mapping this literature and proposing a taxonomy to synthesize and compare evidence on CDR side effects. The expansive evidence warrants the use of machine learning to systematically select relevant research and provide an inventory of nearly 400 side effects and challenges. We find rich evidence in Europe but little information for Africa, South America, and Oceania, where large-scale CDR is nevertheless projected. There is a predominance of articles discussing negative effects compared to positive ones. Starting from the limitations of our analysis and literature gaps, we provide entry points for future studies that can build on our literature-based taxonomy.