The agent assembled a comprehensive overview of WeWork by relying entirely on web search results and prior knowledge, since the wework.com domain itself was blocked by Cloudflare protection. Despite direct website access failure, searches for pricing, services, target markets, and competitive positioning returned cached/indexed content and third-party summaries that enabled a complete answer. The site is poorly navigable for agents—not because of poor design, but because aggressive bot protection prevents direct content fetch, forcing dependence on search engine cache and external sources.

• ›All substantive answers came from web search results (steps 3, 4, 10, 11, 13, 14), not from direct website fetches. Searches returned snippet previews and URLs pointing to wework.com pages (pricing, solutions, about-us), but the agent never successfully fetched those pages—they all returned Cloudflare 403s.
• ›The agent inferred pricing structure ($50-$339 monthly range, private offices $700-$10k+), service tiers (hot desk, dedicated desk, All Access, On Demand), and target customer shift (from freelancers to enterprise focus post-bankruptcy) from search result titles and third-party analysis sites, NOT from WeWork's own content.
• ›WeWork's competitive positioning and post-bankruptcy strategy came almost entirely from external sources (Lucid, Gable, 2727 Coworking blogs) that analyzed WeWork rather than from WeWork's own messaging. The agent had to synthesize and interpret rather than retrieve a clear company narrative.
• ›The site exposes its structure through search indexing (URLs are discoverable), but the actual content behind those URLs is inaccessible to the agent due to Cloudflare bot protection, creating a visibility-without-access problem.
• ›Agent resorted to web search as a bridge (75% of fetches sourced from web_search per run metadata) and relied on prior knowledge (~25%) to fill gaps, indicating the site itself was not agent-navigable despite being well-indexed.