Background: Voltage amplitudes are commonly used for the detection of non-excitable tissue, but may be influenced by the propagation of the activation wavefront with respect to the recording electrode. Objective: The aim of this study was to investigate the influence of changing activation wavefronts on unipolar voltage amplitudes (UnipV) in-silico using tailored models of patients with heart failure. Methods: Five patient-tailored bidomain models were created. Propagating wavefronts and electrograms were computed with state-of-the-art techniques. Simulations were performed with the Propag-5 software on 2304 cores of the Bullx cluster "Curie" (TGCC, CEA, France). The baseline simulation was fitted on geometrical and electro-anatomic mapping data during intrinsic rhythm from heart failure patients. Fibrosis was not incorporated in the simulations allowing only one source of variation on UnipV. UnipV from simulations of single point right ventricular (RV) pacing and left ventricular (LV) pacing were compared at the LV endocardium and epicardium with the baseline simulation (intrinsic rhythm) using paired analyses. Results: A total of 26,872 paired endocardial (5,374±1,165 per patient) and 51,756 epicardial electrograms (10,351±2,068 per patient) were analysed. At baseline, three patients had a left bundle branch block (LBBB) and two patients a non-specific intraventricular conduction defect. The correlation of UnipV between baseline and pacing was poor for both endocardial (RV pacing: R=0.38, LV pacing: R=0.30) as well as epicardial UnipV (RV pacing: R=0.30, LV pacing: R=0.13). The mean absolute change in UnipV between baseline and RV and LV pacing was respectively 3.9±1.2 mV and 3.9±1.2 mV (both 31% of baseline) for the endocardium, and 5.8±5.8 mV and 6.4±5.4 (36% and 41% of baseline) for the epicardium. RV pacing resulted on average in lower endocardial UnipV in 1 patient, and higher UnipV in 4 patients, while lower epicardial UnipV were observed in 4 patients, and higher UnipV in 1 patient (all p<0.001). LV pacing resulted in lower endocardial UnipV in 2 patients and higher UnipV in 3 patients, while lower epicardial UnipV was observed in 4 patients and higher UnipV in 1 patient (all p<0.001). There was no linear correlation between activation time and UnipV, neither for the endocardial nor the epicardial measurements (all R=-0.05-0.05). Simulation bullseye plots of the activation sequence (left panel) and corresponding UnipV distribution (right panel) of the LV endocardium during baseline (in this patient LBBB) and pacing is shown in the figure. Note that the UnipV distribution changes substantially with different activation wavefronts. Conclusion: UnipV’s are strongly influenced by changing activation wavefronts, influencing the characterization of lowvoltage areas and possibly the localization of non-excitable tissue. There is no linear correlation between the activation time and UnipV.