Fluorescent protein-based FRET is a powerful method for visualizing protein-protein interactions and biochemical reactions in living cells. It can be difficult, however, to avoid photobleaching when observing fluorescent cells under the microscope, especially those expressing CFP. We compared the sensitivity of two protein-based FRET pairs to light-induced fluorescence changes in the donor, on FRET determination by fluorescence lifetime imaging microscopy (FLIM). Thanks to the very low excitation light levels of the time- and space-correlated single photon counting (TSCSPC) method, FLIM acquisitions were achieved without donor photobleaching. Here, we show that photobleaching of CFP by a mercury lamp under the microscope induced a decrease in the mean fluorescence lifetime, which interfered with FRET determination between CFP and YFP. Importantly, the range of light-induced variation of the mean fluorescence lifetime of CFP was not proportional to the decrease in the steady state fluorescence intensity and varied from cell to cell. The choice of the CFP/YFP pair therefore requires that the cells be observed and analyzed at very low light levels during the whole FRET experiment. In contrast, the GFP/mCherry pair provided an accurate FRET measurement by FLIM, even if some GFP photobleaching took place. We thus demonstrate that CFP can be an unreliable donor for FRET determination in living cells, due to its photosensitivity properties. We demonstrate that the GFP/mCherry pair is better suited for FRET measurement by FLIM in living cells than the CFP/YFP pair.