Abstract. We describe a formaldehyde (HCHO) retrieval algorithm for the Geostationary Environment Monitoring Spectrometer (GEMS) that will be launched by the Korean Ministry of Environment in 2019. The algorithm comprises three steps: pre-processes, radiance fitting, and post-processes. The pre-processes include a wavelength calibration, and interpolation and convolution of absorption cross-sections; radiance fitting is conducted using a non-linear fitting method referred to as basic optical absorption spectroscopy (BOAS); and post-processes include air mass factor calculations and bias corrections for stripe patterns, background HCHO, and latitudinal biases. In this study, several sensitivity tests are conducted to examine the retrieval uncertainties using the GEMS HCHO algorithm. We evaluate the algorithm with the OMI Level 1B irradiance/radiance data by comparing our retrieved HCHO column densities with OMI HCHO products of the Smithsonian Astrophysical Observatory (OMHCHO) and the Belgian Institute for Space Aeronomy (OMI BIRA). Results show that OMI HCHO slant columns retrieved using the GEMS algorithm are in good agreement with OMHCHO, with correlation coefficients of 0.77–0.91 and regression slopes of 0.92–1.04 for March, June, September, and December 2005. Spatial distributions of HCHO slant columns from the GEMS algorithm are consistent with the OMI BIRA products, but have relatively poorer correlation coefficients of 0.52 to 0.76 compared to those against the OMHCHO products. Also, we compare the satellite results with ground-based MAX-DOAS observations. OMI GEMS HCHO vertical columns are by 0.8–30 % lower than those of MAX-DOAS at Haute-Provence Observatory (OHP) in France, Bremen in Germany, and Xianghe in China. We find that the OMI GEMS retrievals have less bias than the OMHCHO and OMI BIRA products for OHP and Bremen in comparison with MAX-DOAS.