Based on the Toulmin argumentation model and corpus quantitative analysis, this study compares the academic persuasive features of Chinese original, AI-translated, and English original scientific texts. The results show that AI translation can reproduce the basic “claim-evidence” framework but lacks higher-order argumentative elements, with mechanical conversion of logical relations and imbalanced rhetorical strategies. Cross-cultural comparison reveals that Chinese academic writing adopts “explicit persuasion” while English uses “implicit persuasion”, and AI fails to adapt to such differences, reflecting its core shortcoming in argumentative reconstruction rather than language conversion. This study verifies the Toulmin model’s effectiveness in cross-linguistic analysis, provides empirical implications for AI technology optimization, scholar usage strategies, and academic translation industry development, and offers a path to improve academic communication quality in the technological innovation context.