The agent successfully understood PKU by assembling information from web search results and prior knowledge, since the primary domain (pku.edu.cn) and English subdomain failed to load. The agent delivered a comprehensive overview of what PKU does, who it serves, pricing tiers, and competitive positioning versus Tsinghua, but had to work around inaccessible official pages and relied heavily on third-party sources and prior knowledge (33%) rather than native site content.

• ›Step [3] search results surfaced the existence of multiple PKU platforms (english.pku.edu.cn, isd.pku.edu.cn, studyatpku.com), but none of these primary sources loaded successfully; the agent pivoted entirely to web search aggregators and third-party sites.
• ›Steps [8], [9], [10], [11] returned search results that cited specific PDFs (tuition fee schedules from isd.pku.edu.cn) and third-party comparison pages; these search snippets and referenced URLs became the primary sources for tuition, program structure, and competitive differentiation—not the main site itself.
• ›The agent had to explicitly acknowledge website navigation failure (step [1] fetch error, step [5] fetch error, step [6] studyatpku.com loaded but was an online service platform rather than informational content) and rely on 67% web search and 33% prior knowledge rather than native content.
• ›PKU's pricing and program details were sourced from external aggregators (globalscholarships.com, hotcoursesabroad.com) and comparison sites (topuniversities.com, thechinajourney.com) rather than PKU's own published pages, forcing the agent to synthesize from fragmented external sources.
• ›The agent correctly identified confusing gaps: multiple website platforms with unclear primary purpose, variable Chinese language requirements not clearly stated, and tuition prices marked 'to be announced'—all signals that the site does not publish authoritative, machine-readable answers in a unified location.