Industrial ultrasonic tomography (UT) possesses a unique advantage in gas–liquid two-phase distribution measurement and has received broad attention over the past decades. Focusing on a liquid-dominated gas-inclusion distribution reconstruction, we propose the strategy of multi-frequency fusion UT (MFFUT), which comprehensively models the transmission attenuation and self-adaptive fusion of multi-frequency reconstruction images through processing the measurement data with a frequency-dependent forward model and a metric-weighted fusion algorithm. Significant improvements to edge/shape preservation and artifact suppression are observed in both simulative and experimental reconstruction results. Quantitatively, the proposed strategy has a lower relative error and a higher correlation coefficient than any single frequency reconstruction method within all major categories of the reconstruction algorithm and under all inclusion complexities. Accordingly, the MFFUT strategy is proved to provide more accurate gas bubble distribution measurement in industrial processes.