Transcranial magnetic stimulation (TMS)-evoked potentials (TEPs), recorded using electroencephalography (TMS-EEG), offer a powerful tool for measuring causal interactions in the human brain. However, the test-retest reliability of TEPs, critical to their use in clinical biomarker and interventional studies, remains poorly understood. We quantified TEP reliability to: (i) determine the minimal TEP amplitude change which significantly exceeds that associated with simply re-testing, (ii) locate the most reliable scalp regions of interest (ROIs) and TEP peaks, and (iii) determine the minimal number of TEP pulses for achieving reliability. TEPs resulting from stimulation of the left dorsolateral prefrontal cortex were collected on two separate days in sixteen healthy participants. TEP peak amplitudes were compared between alternating trials, split-halves of the same run, two runs five minutes apart and two runs on separate days. Reliability was quantified using concordance correlation coefficient (CCC) and smallest detectable change (SDC). Substantial concordance was achieved in prefrontal electrodes at 40 and 60 ms, centroparietal and left parietal ROIs at 100 ms, and central electrodes at 200 ms. Minimum SDC was found in the same regions and peaks, particularly for the peaks at 100 and 200 ms. CCC, but not SDC, reached optimal values by 60–100 pulses per run with saturation beyond this number, while SDC continued to improve with increased pulse numbers. TEPs were robust and reliable, requiring a relatively small number of trials to achieve stability, and are thus well suited as outcomes in clinical biomarker or interventional studies. •Measured test-retest reliability of TMS-EEG evoked potentials (TEPs).•TEPs achieved substantial concordance when measurements are repeated days apart.•Measurement error is low enough for future group biomarker identification.•Maximum concordance occurs in the N100 and P200 peaks in distinct scalp regions.•Optimal concordance is achieved with 60–100 trials per run.