This study takes the course of Soil Mechanics as a case to examine opportunities and challenges of integrating artificial intelligence (AI) into engineering education. We distributed 326 questionnaires to teachers and students and evaluated nine mainstream large language models (LLMs) on typical Soil Mechanics problems. Results show that 94.48% of respondents support AI application in teaching. All tested models achieve over 85% accuracy in objective questions, but accuracy falls below 65% in calculation problems. These data confirm that AI performs well in knowledge retrieval and objective question solving, yet remains insufficient in complex reasoning tasks. Against traditional teaching limitations such as rigid modes, outdated cases, inefficient administration, and restricted experimental access, we propose four innovative pathways: reforming teaching methods, building digital resource libraries, developing intelligent roll?call systems, and establishing virtual simulation platforms. To address challenges including unreliable AI?generated content, potential reduction in student initiative, and high deployment costs, we recommend strategies such as establishing teacher?student collaboration frameworks, cultivating critical thinking, and promoting resource sharing. These findings provide references for intelligent transformation and high?quality development of engineering education.