Laser-induced breakdown spectroscopy (LIBS) signals in water always suffer strong pulse-to-pulse fluctuations that result in poor stability of the spectrum. In this work, a spectrum normalization method based on acoustic signals measured by a hydrophone immersed in water was developed and compared with laser energy normalization. The characteristics of the acoustic signals were studied first, and the correlations between the acoustic signals and LIBS spectra were analyzed. It showed that the spectral line intensity has a better linear relationship with the acoustic energy than with the laser energy. Consequently, the acoustic normalization exhibited better performance on the reduction of LIBS spectral fluctuation versus laser energy normalization. Calibration curves of Mn, Sr, and Li were then built to assess the analytical performance of the proposed acoustic normalization method. Compared with the original spectral data, the average RSD_C values of all analyte elements were significantly reduced from 5.00% to 3.18%, and the average RSD_P values were reduced from 5.09% to 3.28%, by using the acoustic normalization method. These results suggest that the stability of underwater LIBS can be clearly improved by using acoustic signals for normalization, and acoustic normalization works more efficiently than laser energy normalization. This work provides a simple and cost-effective external acoustic normalization method for underwater LIBS applications.