The Global Ozone Monitoring Experiment (GOME), launched on board the ERS-2 satellite of the European Space Agency, is a spectrometer measuring the Earth's reflectivity between 240 and 790 nm. The main geophysical product of GOME is the ozone vertical column density, also called ozone column. For accurate ozone column retrievals the presence of clouds should be taken into account. Therefore, as part of the operational ozone column retrieval algorithm, cloud fraction is derived by the initial cloud fitting algorithm (ICFA) from the spectral reflectivity between 758 and 778 nm, which encloses the O2 A band. In ICFA, cloud top pressure is assumed a priori, and is taken from the International Satellite Cloud Climatology Project (ISCCP) database. We validated the ICFA (version 2.3) cloud fraction product in two ways. First, a statistical approach was performed comparing monthly average ICFA cloud fractions with monthly average cloud fractions from ISCCP. Global cloud patterns in monthly average ICFA cloud fraction maps compare reasonably well with those from ISCCP. Second, a detailed comparison for individual pixels was performed between ICFA cloud fractions and cloud fractions derived from collocated Along Track Scanning Radiometer-2 (ATSR-2) data. We found that large differences exist between the (effective) cloud fractions from ATSR-2 and ICFA. The mean difference between the cloud fractions of ATSR-2 and ICFA is 0.18; the standard deviation of the difference is 0.23. It is argued that the errors in the ICFA cloud fractions are probably due to errors in the assumed cloud top pressure. A modified version of ICFA, which is less sensitive to the assumed cloud top pressure, is presented. This yields a much improved agreement with the ATSR-2 cloud fractions. Effects of errors in cloud fraction and cloud top height on the vertical ozone column density retrieved from GOME are discussed. ¿ 1999 American Geophysical Union