Abstract: Generative AI has progressively influenced teaching and learning methodologies,fostering innovation in the education sector. Using soil mechanics instruction as a case study,this paper evaluates the pedagogical advantages and limitations of generative AI by analyzing responses from DeepSeek and Gemini through simulated Q&A interactions. The findings indicate that current generative AI demonstrates strong performance in addressing fundamental soil mechanics concepts and solving basic computational problems through formulaic methods. These tools show significant potential for supporting student self-directed learning of core principles. However,constrained by existing technological limitations,generative AI still faces several challenges,including: risks of hallucination; inadequate comprehension of complex soil behaviors; deficiencies in graphical interpretation capabilities; and limited accuracy in solving advanced computational problems. Therefore,in teaching soil mechanics,instructors should prioritize developing students' rational and critical thinking abilities,training them to evaluate AI-generated content rather than accepting it uncritically. Instructional strategies must also emphasize deeper conceptual understanding of complex soil behaviors and strengthen essential computational competencies.