Electromagnetically induced transparency (EIT) -like effects in metamaterials provide a unique approach to enhance the light-matter interactions, and demonstrates great potential in high-sensitivity sensing. However, the conventional optimization methods for terahertz metamaterials rely heavily on the researchers' experiences, so to achieve rapid optimization and high sensitivity across multiple frequency bands remains challenging. To address these challenges, a metamaterial sensor with EIT-like effect is designed and optimized in this work using genetic algorithm, and automatic optimization of parameters is achieved by combining three-dimensional electromagnetic field simulation calculations with genetic algorithms. The results demonstrate that utilizing genetic algorithm to optimize the structural parameters of the sensor can reduce a large number of repetitive manual operations and significantly improve the design efficiency. The designed sensor generates three EIT-like transparent windows in the 0.3–1.5 THz frequency range, and the coupling of the excited electric field with the analyte to be measured is enhanced by means of substrate etching, which significantly improves the sensitivity of the sensor. In addition, in order to verify the potential of the designed EIT metamaterial as a biosensor, simulation-based sensing experiments are conducted for detecting pesticide analytes. The results show that the sensor is able to distinguish substances with different refractive indexes by resonance frequency shift and has the advantage of multi-point feature matching, which has a promising development prospect in the field of sensing and detection.