Gold was determined in ore samples following generation, separation, collection on a graphite cuvette inner wall, and atomization of its volatile species formed by combining an acidified sample solution with an aqueous sodium tetrahydroborate solution at room temperature. A detection limit of 2.6 mug L-1 (3sigma) was obtained with a 5.0 mL sample volume. Precision of replicate measurements was typically 10% RSD. The overall efficiency of the volatile species generation, transport, and trapping process was 0.4%. Atomization of gold from Pd, Ir, Cu, Ag, and W coated graphite cuvettes, following by trapping of the gaseous volatile gold species on these surfaces, has been investigated. Pd coating was found to decrease the sensitivity. Five mug of Ir, 20 mug of Cu, 20 mug of Ag, and 5 mug of W were chosen as optimum masses. W treatment was the best one for sensitivity enhancement, having an improvement factor of 2.4. Silanization of glass surfaces significantly decreased memory effects; improved peak shapes were thus obtained for flow injection vapor generation atomic absorption spectrometry (FI-VGAAS). The method described in this study was used for the determination of gold in an ore reference material, Gold Ore (MA-1b), Canadian Reference Materials Program.