The estimation of an ant’s diet is crucial in many ecological studies. Different techniques, which involve different assumptions and field procedures, have been used to estimate the composition of harvester ant diet. In this study, three techniques are compared for the estimation of the diet of Pogonomyrmex rastratus (Mayr), Pogonomyrmex pronotalis (Santschi), and Pogonomyrmex inermis (Forel) in the central Monte desert, Argentina: (1) hand collection of items brought back to the nest by foragers, (2) collection of items with a semiautomated device with pitfall traps, and (3) collection of the discarded material accumulated in middens. The hand collection technique and the collection of middens provided the lowest and the highest number of items, respectively. Midden samples and pitfall traps contained a higher proportion of nonseed items, probably coming from sources other than ants, than hand-collected items. The three techniques provided similar estimations of species richness but a bias against small seeds was detected for P. pronotalis and P. inermis with the hand collection technique, possibly because of the difficulty of collecting small items by hand. The percentage of seed species in the diet obtained with different techniques was positively correlated in the great majority of colonies. Overall, despite their intrinsic differences, the three techniques proved consistent, which constitutes a robustness test for the estimations obtained. In comparative ecological studies, the awareness that results depend on the techniques and their assumptions is particularly important.