The study aimed to investigate the accuracy of diagnosing thin and thick gingival phenotypes (GP) by the transparency of the periodontal probe (TRAN) through the gingival sulcus. Eligible studies comparing TRAN to direct methods for gingival thickness (GT) measurement (reference tests) were searched in 4 databases (including MEDLINE and EMBASE), up to 2024. Quality assessment was carried out using QUADAS-2. Latent class meta-analysis for imperfect gold standard model was conducted considering the multiple thresholds (TSs) and landmarks adopted by the reference tests. The 10 studies included presented a low risk of bias and low applicability concerns. The summary sensitivity (SSe) ranged from 49% (95% CI:25.8-68%) with the TS 0.8mm, to approximately 53% (TSs 1 and 1.2mm); the summary specificity (SSp) ranged from 60% (95%CI: 42.4-82.6%) with the TS 1mm, to 68% (95% CI: 46.4-82.5%) with the TS 0.8mm. The highest SSe (67%) and SSp (76%) were found in the analysis grouping the same TS (0.8mm) and landmark (1mm from gingival margin). The assessment of GP using the TRAN could lead to inadequate diagnoses, especially in thin phenotype determination. Its accuracy is highly dependent on the TSs used to differentiate between thin and thick GPs and the apicocoronal landmark where the GT was obtained.