Background: Potentially inappropriate prescribing (PIP) is frequent in geriatrics and results in an increased risk for adverse effects, morbidity, mortality and reduced quality of life. Research on PIP in psychiatry has mainly focused on elderly patients and inpatients. Objectives: To determine the prevalence and the predictors of PIP of psychotropic medication in outpatients with severe mental illness. Design: This study is part of the Muva study, a pragmatic open Stepped Wedge Cluster Randomized Trial of a physical activity intervention for patients (age ⩾ 16 years) with severe mental illness. Methods: A structured medication interview, questionnaires on social functioning, quality of life and psychiatric symptoms, and BMI and waist circumference measurements were performed followed by a structured medication review. Patients were divided into groups: PIP versus no PIP. Between-group differences were calculated and a multivariate binary logistic regression was performed to examine predictors for PIP. A receiver operating characteristics analysis was performed to determine the area under the curve (AUC). Results: In 75 patients, an average of 5.2 medications of which 2.5 psychotropic medication was used. 35 (46.7%) patients were identified with PIP. Unindicated long-term benzodiazepine use was the most frequently occurring PIP (34.1%). Predictors of PIP were female gender [odds ratio (OR) = 4.88, confidence interval (CI) = 1.16–20.58, p = 0.03], number of medications (OR = 1.41, CI = 1.07–1.86, p = 0.02) and lower social functioning (OR = 1.42, CI = 1.01–2.00, p = 0.05). The AUC was 0.88 for the combined prediction model. Conclusion: The prevalence of PIP of psychotropic medication in outpatients with severe mental illness is high. It is therefore important to identify, and where possible, resolve PIP by frequently performing a medication review with specific attention to females, patients with a higher number of medications and patients with lower social functioning. Trial registration: This trial was registered in The Netherlands Trial Register (NTR) as NTR NL9163 on 20 December 2020 ( https://trialsearch.who.int/Trial2.aspx?TrialID=NL9163 ).