In March 2002 the European Space Agency (ESA) launched the polar-orbiting environmental satellite Envisat. One of its nine instruments is the Global Ozone Monitoring by Occultation of Stars (GOMOS) instrument, which is a medium-resolution stellar occultation spectrometer measuring vertical profiles of ozone. In the first year after launch a large group of scientists performed additional measurements and validation activities to assess the quality of Envisat observations. In this paper, we present validation results of GOMOS ozone profiles from comparisons to microwave radiometer, balloon ozonesonde, and lidar measurements worldwide. Thirty-one instruments/launch sites at twenty-five stations ranging from the Arctic to the Antarctic joined in this activity. We identified 6747 collocated observations that were performed within an 800-km radius and a maximum 20-hour time difference of a satellite observation, for the period between 1 July 2002 and 1 April 2003. The GOMOS data analyzed here have been generated with a prototype processor that corresponds to version 4.02 of the operational GOMOS processor. The GOMOS data initially contained many obviously unrealistic values, most of which were successfully removed by imposing data quality criteria. Analyzing the effect of these criteria indicated, among other things, that for some specific stars, only less than 10% of their occultations yield an acceptable profile. The total number of useful collocated observations was reduced to 2502 because of GOMOS data unavailability, the imposed data quality criteria, and lack of altitude overlap. These collocated profiles were compared, and the results were analyzed for possible dependencies on several geophysical (e.g., latitude) and GOMOS observational (e.g., star characteristics) parameters. We find that GOMOS data quality is strongly dependent on the illumination of the limb through which the star is observed. Data measured under bright limb conditions, and to a certain extent also in twilight limb, should be used with caution, as their usability is doubtful. In dark limb the GOMOS data agree very well with the correlative data, and between 14- and 64-km altitude their differences only show a small (2.5--7.5%) insignificant negative bias with a standard deviation of 11--16% (19--63 km). This conclusion was demonstrated to be independent of the star temperature and magnitude and the latitudinal region of the GOMOS observation, with the exception of a slightly larger bias in the polar regions at altitudes between 35 and 45 km.