To analyse, by means of a meta-analytical approach, the diagnostic accuracy of molecular biomarkers in gingival crevicular fluid (GCF) for the detection of periodontitis in systemically healthy subjects. Studies on GCF molecular biomarkers providing a binary classification table (or sensitivity and specificity values and group sample sizes) in individuals with clinically diagnosed periodontitis were considered eligible. The search was performed using six electronic databases. The methodological quality of studies was assessed through the tool Quality Assessment of Diagnostic Studies. Meta-analyses were performed using the Hierarchical Summary Receiver Operating Characteristic, which adjusts classification data using random effects logistic regression. The included papers identified 36 potential biomarkers for the detection of periodontitis and for four of them meta-analyses were performed. The median sensitivity and specificity were for MMP8, 76.7% and 92.0%; for elastase, 74.6% and 81.1%; for cathepsin, 72.8% and 67.3%, respectively. The worst estimates of sensitivity and specificity were for trypsin (71.3% and 66.1%, respectively). MMP8 showed good sensitivity and excellent specificity, which resulted in this biomarker being clinically the most useful or effective for the diagnosis of periodontitis in systemically healthy subjects, regardless of smoking condition. © 2019 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd.