Thanks to their enormous energy release which allows to detect them up to very high redshift, Gamma Rays Bursts (GRBs) have recently attracted a lot of interest to probe the Hubble diagram (HD) deep into the matter dominated era and hence complement Type Ia Supernoave (SNeIa). However, lacking a local GRBs sample, calibrating the scaling relations proposed as an equivalent to the Phillips law to standardize GRBs is not an easy task because of the need to estimate the GRBs luminosity distance in a model independent way. We consider here three different calibration methods based on the use of a fiducial $\Lambda$CDM model, on cosmographic parameters and on the local regression on SNeIa. We find that the calibration coefficients and the intrinsic scatter do not significantly depend on the adopted calibration procedure. We then investigate the evolution of these parameters with the redshift finding no statistically motivated improvement in the likelihood so that the no evolution assumption is actually a well founded working hypothesis. Under this assumption, we then consider possible systematics effects on the HDs introduced by the calibration method, the averaging procedure and the homogeneity of the sample arguing against any significant bias. We nevertheless stress that a larger GRBs sample with smaller uncertainties is needed to definitely conclude that the different systematics considered here have indeed a negligible impact on the HDs thus strengthening the use of GRBs as cosmological tools.