Importance: The concept of embolic stroke of undetermined source (ESUS) unifies a subgroup of cryptogenic strokes based on neuroimaging, a defined minimum set of diagnostic tests, and exclusion of certain causes. Despite an annual stroke recurrence rate of 5%, little is known about the etiology underlying recurrent stroke after ESUS. Objective: To identify the stroke subtype of recurrent ischemic strokes after ESUS, to explore the interaction with treatment assignment in each category, and to examine the consistency of cerebral location of qualifying ESUS and recurrent ischemic stroke. Design, Setting, and Participants: The NAVIGATE-ESUS trial was a randomized clinical trial conducted from December 23, 2014, to October 5, 2017. The trial compared the efficacy and safety of rivaroxaban and aspirin in patients with recent ESUS (n = 7213). Ischemic stroke was validated in 309 of the 7213 patients by adjudicators blinded to treatment assignment and classified by local investigators into the categories ESUS or non-ESUS (ie, cardioembolic, atherosclerotic, lacunar, other determined cause, or insufficient testing). Five patients with recurrent strokes that could not be defined as ischemic or hemorrhagic in absence of neuroimaging or autopsy were excluded. Data for this secondary post hoc analysis were analyzed from March to June 2019. Interventions: Patients were randomly assigned to receive rivaroxaban, 15 mg/d, or aspirin, 100 mg/d. Main Outcomes and Measures: Association of recurrent ESUS with stroke characteristics. Results: A total of 309 patients (205 men [66%]; mean [SD] age, 68  years) had ischemic stroke identified during the median follow-up of 11 (interquartile range [IQR], 12) months (annualized rate, 4.6%). Diagnostic testing was insufficient for etiological classification in 39 patients (13%). Of 270 classifiable ischemic strokes, 156 (58%) were ESUS and 114 (42%) were non-ESUS (37 [32%] cardioembolic, 26 [23%] atherosclerotic, 35 [31%] lacunar, and 16 [14%] other determined cause). Atrial fibrillation was found in 27 patients (9%) with recurrent ischemic stroke and was associated with higher morbidity (median change in modified Rankin scale score 2 [IQR, 3] vs 0 (IQR, 1]) and mortality (15% vs 1%) than other causes. Risk of recurrence did not differ significantly by subtype between treatment groups. For both the qualifying and recurrent strokes, location of infarct was more often in the left (46% and 54%, respectively) than right hemisphere (40% and 37%, respectively) or brainstem or cerebellum (14% and 9%, respectively). Conclusions and Relevance: In this secondary analysis of randomized clinical trial data, most recurrent strokes after ESUS were embolic and of undetermined source. Recurrences associated with atrial fibrillation were a minority but were more often disabling and fatal. More extensive investigation to identify the embolic source is important toward an effective antithrombotic strategy. Trial Registration: ClinicalTrials.gov Identifier: NCT02313909..