Before Gaia data are published, they will undergo scrutiny in several different validation processes. Each leg of the data reduction will have its own specialised validation with focus on the difficult points for that task. In addition, when the combined catalogue is ready for release, it will undergo an independent, more global, validation. All kinds of properties of the catalogue are checked, and it is of particular interest to test if the error estimates are realistic, and if the parameters are unbiased. We also look briefly at the experiences from other projects like 2MASS and Hipparcos, and address recent worries that Hipparcos was wrong and that Gaia may therefore also be wrong.