The agent successfully assembled a comprehensive explanation of dndn by relying almost entirely on web search results and prior knowledge, since the dndn.io website itself returned only JavaScript-heavy shells with minimal accessible content. The agent pieced together what dndn does (Korean robo-advisor for retirement), who it serves (Korean investors with pension accounts), its 0.3% fee structure, and differentiators (pension focus, global diversification, founder credibility) — but had to work around the site's poor machine-readability and fill gaps using external sources like Namu.wiki, Korea Times articles, and app store listings.

• ›Direct fetches of dndn.io, product.dndn.io, and dndn.io/aboutus all returned identical truncated HTML with meta tags and Google Tag Manager boilerplate, but no actual product content — all real content is rendered client-side via JavaScript and inaccessible to text-based fetches.
• ›The agent's final answer was sourced primarily from web search results (steps 7, 11, 17, 25, 31) which surfaced third-party sources (Namu.wiki, Korea Times, app store listings, Crunchbase) rather than dndn's own website; the dndn.io domain itself provided zero substantive information.
• ›Critical details like the 0.3% fee, merger consolidation in late 2024, founder Kim Sung-il's background, and all-weather portfolio philosophy came from search results and external sources, not from dndn's official site — indicating dndn publishes this information nowhere fetchable or indexable.
• ›The agent explicitly called out major gaps (English language support, detailed product breakdown, fee variations, exact competitive positioning, brand name clarity) that exist because dndn's website does not expose this information in machine-readable form.
• ›The site's own content was so inaccessible that the agent's prior knowledge (29%) and web search routing (57%) combined to carry the task; the final answer cites external sources, not dndn's domain.