The assessment of safety data has become a standard across many clinical interventions. The aim of this systematic review is to investigate the extent to which harm is addressed within psychotherapy study protocols. The review includes study protocols of randomized controlled trials published between 2004 and 2017 investigating the effects of psychotherapy in adult patients with affective disorders, phobia, anxiety, obsessive-compulsive disorder, posttraumatic stress disorder, and/or personality disorders. We conducted a systematic search in the CENTRAL, Medline, PsycINFO, and Web of Science databases as well as in relevant journals. In total, 115 study protocols were included, examining 168 psychotherapy and 85 control conditions. These protocols differed considerably in the way they conceptualized harm: 77 explicitly addressed harm, 62 considered serious adverse events, and 39 considered adverse events. Although serious adverse events were defined somewhat consistently, adverse events were not. Our results imply that clinical researchers do not apply standardized approaches with regard to harm concepts, assessment, and management. To gather data on frequencies of harmful effects, we argue a higher degree of standardization would be useful. Feasible recommendations are provided based on examples Bernhard Strauss contributed equally to supervision and served as support for methodology, validation, original draft, and writing, review, and editing. Christoph Flückiger served as support for data curation and writing, review, and editing. Francesca Färber served as lead for methodology and project administration, contributed equally to conceptualization, funding acquisition, and supervision, and served as support for data curation, formal analysis, and investigation. Jenny Rosendahl served as editor.