The agent successfully assembled a comprehensive overview of Jana Vidya Foundation by combining direct site fetches with targeted web searches to fill gaps. The site itself is navigable but content-sparse in rendered HTML—most pages are React/Next.js shells that don't expose text to fetchers—forcing heavy reliance on web search results and prior knowledge (47% of fetches) to answer questions about programs, pricing, impact, and differentiation. The agent delivered a satisfying answer despite the site's limited machine-readability.

• ›Content-bearing pages ([0] homepage, [1] about, [4] donate, [5] contact, [9] get-involved, [19] our-work, [20] transparency, [21] scholarships) returned only React bundle shells with minimal static text, so the agent could not extract structured details about programs, scholarship amounts, donation tiers, or eligibility criteria directly from HTML.
• ›The sitemap ([18]) was the agent's only successful structured discovery mechanism, guiding it to /our-work/, /transparency/, and /scholarships/ after initial guesses (/services, /pricing, /programs, /work) returned 404s, indicating poor URL predictability.
• ›Web searches ([8], [10], [11], [12], [14], [15], [16]) provided the bulk of factual assertions in the final answer—founder names, program counts, volunteer numbers, the '₹70 per ₹100' spend ratio, founding year 2023, and geographic focus—none of which the agent could extract from the site's fetched HTML.
• ›The agent explicitly acknowledged missing details (donation tiers, detailed eligibility criteria, outcome metrics, capacity/waitlist info) and correctly attributed these gaps to the site's client-side rendering architecture rather than claiming they don't exist.
• ›The agent cited five URLs in its sources, but only [0], [1], [4], [9], [19] were actually fetched and returned 200s with meaningful structure; [5] and [20] (contact, transparency) were fetched but their content was not extracted.