Electron multiplying charge-coupled devices (EMCCDs) are a variant of standard CCD technology capable of single-optical photon counting at MHz pixel readout rates. For photon counting, thermal dark signal and clock-induced charge (CIC) are the dominant source of noise and must be minimized to reduce the likelihood of coincident events. Thermal dark signal is reduced to low levels through cooling or operation in inverted mode (pinning). However, mitigation of CIC requires precise tuning of both parallel and serial clock waveforms. Here, we present a detailed study of CIC within Teledyne-e2v EMCCDs with a goal of better understanding the physical mechanisms that dominate CIC production in both noninverted and inverted mode operations (IMO). Measurements are presented as a function of parallel and serial clock timings, clock amplitudes, and device temperature. The effects of radiation damage and annealing are also discussed. A widely accepted view is that CIC is signal generated through impact ionization of energetic holes as the clock phase is driven high. While this explanation holds for IMO, we propose that the majority of CIC generated in noninverted mode is in fact due to a secondary effect of light emission from hot carriers. The information from this study is then used to optimize CIC on Teledyne e2v CCD201s operating at 1-MHz pixel rate in NIMO. For the CCD201, we obtained total CIC levels as low as 6.9  ×  10^−  4  e⁻  /  pix  /  frame with ≥90  %   detective quantum efficiency. We conclude with proposals to further reduce CIC based upon modifications to clocking schemes and device architecture.