Global ozone profiles are derived from the ultraviolet and visible part of the spectra of the nadir-viewing Global Ozone Monitoring Experiment (GOME), which is mounted on the polar orbiting second Earth Remote Sensing satellite (ERS-2). These profiles need to be characterized, especially since the product includes a priori knowledge and so-called averaging kernels. This additional information needs to be taken into account when comparing the profiles to correlative measurements. We perform an intercomparison between the ground-based stratospheric lidar system in Lauder, New Zealand, and collocated GOME data. Here, the satellite profiles are retrieved with the algorithm of the Royal Netherlands Meteorological Institute (KNMI), which uses the optimal estimation method. In the comparison study significant differences are revealed which vary with season and altitude, indicating errors in the retrieval system. However, any quality assessment will just be one part of characterizing an ozone profile product that includes averaging kernels and a priori information. Data users need to be aware of the inherently complicated nature of such products that can only be fully understood when taking into account this additional information. In the second part of the study, the complex relation between the retrieved and the true profile is clarified using several interpretation tools. Applying these tools, we conclude that below 17-km altitude the GOME profiles can only be used with appropriate use of averaging kernels (e.g., in data assimilation). Above 17 km up to 50 km the GOME spectra contain useful profile information, but the retrieved profiles have a moderate vertical resolution of about 11 km and contain a substantial fraction of a priori information of about 50%.