生成式AI能够快速生成词义解释与教学例句，但其近义词辨析输出能否直接用于国际中文教学，仍需结合具体语言材料考察。本文以“了解、理解、明白”为对象，从语义、句法与语用三个层面建立分析框架，对豆包网页端快速模式开展探索性测试。15道人工编制题在三个新会话中重复提交，共取得45个题目回答单元。定性分析发现，模型对典型语境的优选词及若干不自然搭配给出了较稳定的核心判断，但部分回答在例句功能展示、解释边界及替代表达接受度方面存在待复核问题。三轮核心判断相近并不意味着解释完全一致，也不能证明教学有效。本文据此提出“语境比较—解释核查—句式改写—交际运用”的教学活动方案，强调教师对生成材料的筛选与修订。研究结论仅适用于本次单产品、小样本及指定提示词条件。

Generative AI can produce explanations and examples for vocabulary teaching, but the suitability of its near-synonym distinctions requires close examination. This exploratory study examines “了解”, “理解” and “明白” through semantic, syntactic and pragmatic analysis. Fifteen constructed tasks were submitted as a complete set in each of three new conversations using Doubao’s web interface in its visible quick mode, producing 45 task-level responses from three requests. Qualitative analysis found relatively stable core judgments in typical contexts, alongside issues requiring review in example functions, explanation boundaries and the acceptability of alternatives. Similar core judgments did not imply identical explanations or demonstrated teaching effectiveness. A classroom activity involving contextual comparison, explanation checking, sentence rewriting and communicative use is proposed. The findings apply only to this product, small task set and specified prompting condition; independent teacher review remains necessary.