The agent assembled a functional overview of EPITA by relying almost entirely on web search results and third-party education sites, because the epita.fr domain itself delivered truncated, unreadable HTML. The final answer is reasonably complete—covering what EPITA is, who it serves, approximate pricing, and competitive differentiation—but is explicitly grounded in external sources rather than direct site content, and the agent called out significant gaps in verified detail (current programs, admission requirements, scholarship specifics, career outcomes) that could not be accessed from the primary domain.

• ›Steps [1], [2], [4], [7], [8] attempted to fetch epita.fr directly but returned compressed/truncated HTML that was not machine-readable; the agent correctly recognized this as a blocker and pivoted to web search.
• ›All four content-bearing steps ([6], [10], [12], [14]) were web searches that surfaced third-party aggregator sites (study.eu, shiksha.com, chunocollege.com, jeduka.com) and EPITA's own URLs embedded in search snippets—not the main domain itself. The agent cited these search results' summaries and the URLs they contained as sources.
• ›The agent's pricing figures, program descriptions, and competitive comparisons came from third-party education platforms (Shiksha, Study.eu, UniPage, Kadamb Overseas), not from epita.fr's own pricing or admissions pages. The website did not publish this information in a form the agent could retrieve.
• ›While the agent discovered EPITA's existence, positioning (AI/cybersecurity focus), and general structure through search, it explicitly flagged that the admissions page, program curriculum, scholarship details, and career outcomes data remain unverified because they could not be fetched from the primary domain.
• ›The final response's source list contains 12 URLs, of which only 3 are epita.fr domains; 9 are third-party aggregators or comparison sites, indicating heavy reliance on secondary sources to construct an answer that should have come from primary sources.