A long-range geoid testing procedure is presented using a geometrical geoid profile, which extends in an approximate north-south direction between the Mediterranean and the Black Sea. Along the first-order leveling line of the Turkish National Vertical Control Network (TNVCN), 47 leveling benchmarks at 20-25 kin intervals were selected as global positioning satellite (GPS)-leveling control points with known Helmert orthometric heights. The geometrical geoid profile computed from GPS-leveling data was compared with the national Turkish Geoids 1991 (TG91) and 1999A (TG99A), and global geoids OSU91 A and EGM96. In absolute geoid testing, it is verified that TG91 has a tilt and bias with respect to the GPS-leveling geoid. TG99A, which has no tilt and bias, displays the root mean square agreement of +/- 12 cm in the north-south direction. In relative geoid testing, it is understood that TG91 can only be used at 25 km or shorter baseline lengths, whereas TG99A can be used over all baseline lengths with an accuracy that is equal to the tolerance value for the second-order leveling control.