Sinusoidal pulse width modulation (SPWM) is a conventional control technique for three-phase voltage source inverters (VSIs), but it often introduces electromagnetic interference (EMI) and acoustic noise in connected induction loads. Random pulse width modulation (RPWM) offers better EMI performance but suffers from inefficient DC bus utilisation and residual harmonic clusters. This study proposes an artificial neural network (ANN)-optimised RPWM technique to more effectively disperse harmonic energy across a wider frequency spectrum. The ANN-generated modulation signal was integrated into a three-phase VSI with a passive L(Inductor), C(Capacitor) (LC) filter and evaluated against dual random and fixed PWM strategies in MATLAB/Simulink (Mathworks MATLAB R2023a) using power spectral density (PSD) and total harmonic distortion (THD) analyses. Results show that the ANN-based RPWM reduces conducted EMI and achieves a THD of 2.17%, with only a negligible increase in computational cost and no additional hardware requirements, offering a practical and cost-efficient approach for EMI reduction in inverter systems.