Introduction: Our ability to determine the pathogenicity of variants is limited by the paucity of clinical data when variants are seen in very few individuals. In vitro functional assays could help overcome the deluge of variants of uncertain significance (VUS). To be clinically useful, such assays must satisfy stringent benchmarking criteria established by the ACMG/AMP. For variants in KCNH2 , the cause of Long QT Syndrome Type 2 (LQTS2), ~70% of missense variants are VUS. Here, we wish to formally evaluate the accuracy, replicability and robustness of a high-throughput automated patch-clamp (APC) assay for characterising KCNH2 variants. Aim: To evaluate an APC assay for KCNH2 variants against the criteria for the application of functional evidence within the ACMG/AMP framework. Methods: Twenty five variants classified as pathogenic/likely pathogenic and 18 variants classified as benign/likely benign from ClinVar were heterozygously expressed in stable Flp-In HEK293 cells and analysed for loss-of-function based on current density and changes in channel gating. Results: Twenty three of 25 pathogenic variants had <50% wild-type (WT) peak current density, indicating loss-of-function. The two remaining pathogenic variants, G584S and N629S, had <50% WT current levels (IKr) after incorporating additional gating measures. All 18 benign variants had >50% WT current levels both before and after including gating measures. For 17 of the benign variants, current levels were not statistically different to WT. The current level for R148W was significantly different to WT (p<0.005) but was still >50%. Based on a 50% cut-off, the assay yields a sensitivity of 100% and specificity of 100%. Conclusions: High-throughput electrophysiological phenotyping correctly predicted loss of function for 25/25 pathogenic variants, and 18/18 benign variants. The assay should be able to assist with the clinical classification of KCNH2 variants.