The agent successfully assembled a comprehensive explanation of QwenCloud by combining content from the homepage, pricing page, models marketplace, and documentation—however, it relied heavily on web search and prior knowledge to work around a largely non-navigable site structure. The site's URL routing is broken (most guessed paths returned 404s), forcing the agent to discover correct URLs via search results rather than through site navigation or information architecture.

• ›Step [1] (homepage) provided the title and meta keywords but no actionable navigation paths—the HTML was truncated and did not contain visible links to pricing or products sections. Steps [3, 4, 6, 7] all returned 404s when the agent attempted logical URL paths (/pricing, /products, /en, /model), indicating the site does not expose its information hierarchy through standard routes.
• ›Steps [11] (pricing page) and [12] (models marketplace) were only discovered via web search [9], not by following links from the homepage. This means QwenCloud's core content—what it costs and what models it offers—is not internally linked or discoverable by navigation alone. The agent had to use external search to find www.qwencloud.com/pricing/api and www.qwencloud.com/models.
• ›Step [17] (documentation at docs.qwencloud.com) was also discovered via search, not site navigation. This indicates the main site and docs are siloed and not cross-linked in a way agents can traverse. The agent assembled the final answer by stitching together fragments from multiple sources (homepage metadata, search-discovered pricing/models pages, external third-party blogs, and GitHub documentation) rather than following a coherent site-guided journey.
• ›The site's agent-readiness is poor: meta tags and title exist (good for SEO), but the actual navigable structure is broken or absent. A human visiting qwencloud.com would likely see a working interface (JavaScript-rendered), but the agent's fetch-based approach revealed that either the HTML structure lacks proper links, or key pages exist only at non-obvious URLs. The agent's 'heavy_bridge' friction classification is justified—it required search queries and prior knowledge to compensate for missing internal navigation.