Background: There is conflicting evidence about the use of biomarkers to diagnose left atrial thrombus in patients with atrial fibrillation. Objective: The purpose of this study was to assess the diagnostic accuracy of D-dimer to detect left atrial thrombus in patients with atrial fibrillation. Methods: We searched 4 electronic databases from inception to December 16, 2020. The reference standard was left atrial thrombus detected by transesophageal echocardiography. Study quality was assessed with the Quality Assessment of Diagnostic Accuracy Studies 2 tool. We used a bivariate model to calculate the pooled sensitivity and specificity with their 95% confidence intervals (CIs). The optimal cutoff and predictive values were also estimated. Results: Eleven cross-sectional studies involving 4380 patients were included. The median prevalence of left atrial thrombus was 12%. In 7 studies, the pooled sensitivity of D-dimer at 500 ng/mL was 50% (95% CI 26%–74%) and the pooled specificity was 88% (95% CI 76%–95%). The pooled sensitivity of age-adjusted D-dimer was 36% (95% CI 14%–66%) and the pooled specificity was 99% (95% CI 96%–99%) in 2 studies. The optimal cutoff of D-dimer was 390 ng/mL in 10 studies with a pooled sensitivity of 68% (95% CI 44%–85%) and a pooled specificity of 73% (95% CI 54%–86%). The positive and negative predictive values were 21.8% and 95.4%, respectively. The risk of bias was low or unclear for all domains. Concerns about applicability were low for almost all studies. Conclusion: Our meta-analysis suggests that D-dimer has the potential to be useful to rule out left atrial thrombus in patients with atrial fibrillation.