Requirements elicitation interviews are crucial in requirements engineering but still heavily rely on experienced requirements analysts. Recent advancements in large language models (LLMs) have created new opportunities for automating this process. However, relying solely on LLMs for free-form conversations provides limited support for structured interviews, often leading to redundant questions and missed implicit requirements. By taking website applications (i.e., WebApp) as an example, this paper proposes TypoAgent, a structured requirements interview agent. By analyzing WebApp requirements artifacts, TypoAgent automatically constructs a reusable experience typology tree, which organizes requirement dimensions into a hierarchical inquiry space to support systematic interview planning. Guided by this typology tree, TypoAgent further applies initial priority ranking to align with the user’s underspecified initial intent, and adopts context-aware dynamic reranking to adaptively adjust questioning directions during multi-turn interactions. In addition, TypoAgent introduces a pruning mechanism to reduce ineffective questions and improve interview efficiency. Experimental results demonstrate that TypoAgent significantly outperforms existing baselines in both implicit requirement elicitation effectiveness and questioning efficiency. Moreover, ablation studies confirm the effectiveness of each key design choice in TypoAgent. We believe that TypoAgent is highly transferable and can be extended to various domain conversational requirements elicitation scenarios.