Background: Detection of atrial fibrillation after ischemic stroke is challenging due to its paroxysmal nature. We aimed to assess predictors of paroxysmal atrial fibrillation using non-invasive surface ECG and transthoracic echocardiography to select candidates for atrial fibrillation screening. Methods: Ischemic stroke patients without documented atrial fibrillation (n = 110, 67 ± 10 years, 40 female) and a control group of age- and gender-matched patients with history of paroxysmal atrial fibrillation prior to stroke (n = 55, 67 ± 10 years, 19 female) comprised the study sample. Using non-invasive ECG monitoring for three weeks, short episodes of paroxysmal atrial fibrillation were detected in 24 of 110 patients (22 %). The standard 12-lead ECG with sinus rhythm at stroke onset was digitally processed and analyzed. Transthoracic echocardiography data were reviewed for these patients. Results: Atrial fibrillation history was independently associated with P terminal force in lead V 1 > 40 mm*ms (OR 4.04 95 % CI 1.34-12.14, p = 0.013) and left atrial volume index (OR 1.08 95 % CI 1.03-1.13, p = 0.002; for LAVI > 40 mL/m² OR 6.40 95 % CL 1.47-27.91, p = 0.013). Among patients without atrial fibrillation history, no ECG characteristics were predictive of atrial fibrillation detected after stroke. Left atrial volume index remained an independent predictor of atrial fibrillation detected after stroke (OR 1.09 95 % CI 1.02-1.16, p = 0.017). A cutoff of <40 mL/m² had an 84 % negative predictive value for ruling out atrial fibrillation on ambulatory monitoring with a sensitivity of 50 % and a specificity of 86 %. Conclusion: In a post hoc analysis, left atrial dilatation assessed by left atrial volume index independently predicted atrial fibrillation after stroke in patients without prior atrial fibrillation history, while the other clinical or ECG markers were not predictive of atrial fibrillation detected early after ischemic stroke. Trial registration: This study is a post hoc analysis from the prospective case-control study registered in December 2011, ClinicalTrials.gov ID: NCT01325545.