This paper presents an exhaustive evaluation of the quality of an audio watermarking algorithm. The integration of the psychoacoustic model into the audio watermarking approach is demonstrated. The quality parameter relating the power of the watermark noise and the masking threshold is presented. The evaluation method is detailed and the quality of the watermarked audio tracks is evaluated with regard to different settings of the quality parameter used to adjust the power of the embedded watermarks. The subjective listener test compares the quality of the original audio track with the watermarked one. Different quality parameter settings were used in order to enable the adjustment between quality and maximum robustness according to the items to be watermarked and the target audience.