Characterizations of the tinnitus sound percept are always based on a subjective description by the person affected. Since the experimenter cannot have access to the tinnitus percept, it is not possible to verify whether individuals use the adjectives describing the sound in the expected way, i.e., whether a label given to the tinnitus percept corresponds to the label that the experimenter or another individual would give to the same sound percept. However, if it is assumed that tinnitus patients can reliably describe their own tinnitus, then they should also be able to reliably describe tinnitus-like sounds, presented acoustically. In this study, 26 tinnitus patients used a tablet computer to rate 18 pre-defined adjectives on their level of descriptiveness for their own tinnitus percept as well as 17 tinnitus-like sounds presented via headphones. The main interest of the current study was to calculate intraclass correlation (ICC) and Krippendorff's alpha coefficients for the rating profiles of the acoustically-presented sounds, in order to quantify how well the individuals agreed on the ratings of known sounds, i.e., whether the adjectives would receive similar ratings from all participants for a specific tinnitus-like sound. The results show that the level of agreement was low for all adjectives and sounds, meaning that the different individuals did not use the adjectives in a consistent manner. The conclusion is that subjective tinnitus descriptions should be interpreted with great caution, and that the inherent variability involved in the characterization of sounds by naïve listeners can contribute to the observed heterogeneity in tinnitus symptoms and treatment outcomes.