The agent could not access Insper's main website directly (403 errors) and instead assembled a partial profile from web search results and third-party sources. It successfully identified what Insper teaches, who attends, and key differentiators, but could not retrieve specific pricing, admission requirements, or detailed program catalogs.

• ›The target domain (insper.edu.br) was inaccessible via direct fetch (403 errors on steps [0] and [2]), forcing reliance entirely on web search results and external sources (Wikipedia, TopUniversities, SKEMA, etc.).
• ›Search results supplied high-level institutional facts (founding, mission, program categories, accreditations, rankings) but no detailed pricing or admission criteria—gaps the agent correctly flagged as missing rather than inferring.
• ›The agent was forced to cite generic third-party education comparison sites rather than Insper's own official pages, weakening authority for specific claims about programs and differentiators.
• ›No site-native content (blogs, PDFs, structured data) was successfully retrieved to ground the answer; all substance came from search snippets and external aggregators, indicating poor bot-readability or access restrictions on the primary domain.
• ›The agent accurately acknowledged the limits of what it could retrieve and did not hallucinate missing facts (e.g., specific tuition amounts or admission test scores), demonstrating appropriate epistemic honesty.