Passive satellite retrievals using conventional CO2 absorption techniques tend to systematically underestimate the upper transmissive cloud top heights (CTHs). These techniques are based on single‐layer assumptions that the upper cloud occupies a geometrically thin layer above a cloud‐free surface. This study presents a new modified CO2 absorption technique (MCO2AT) to improve the inference of transmissive CTHs in the upper troposphere above 600 hPa. The MCO2AT employs an iterative algorithm that starts with a single‐layer CO2 absorption technique (SCO2AT) followed by an iterative procedure to retrieve an enhanced upper CTH based on inferred effective background radiances. Both techniques are applied to the 10.7 and 13.3 mm channel data of the Twelfth Geostationary Operational Environmental Satellite (GOES 12) imager and their retrievals of upper tropospheric CTHs are compared with two active sensing products: the ground‐based Active Remotely Sensed Cloud Location (ARSCL) products from the Atmospheric Radiation Measurement Program (ARM) Southern Great Plains (SGP) site and the satellite‐based Cloud Aerosol Lidar With Orthogonal Polarization (CALIOP) products. On average, the CTHs from MCO2AT and SCO2AT are lower than those from both of the active sensors by ∼1 and 2.4 km, respectively, possibly due to the different sensitivities and spatial resolutions between passive and active sensors. Preliminary validation of the new modified method is encouraging, especially the improvements for upper transmissive clouds in geometrically thick and/or multilayered cloud situations. The development of the modified method is particularly useful for sensors like the GOES 12, Meteosat‐9, and others, which carry only one CO2 absorption channel at ∼13.3 mm.