The agent successfully assembled a comprehensive overview of CELTESTPIP by combining direct site fetches with web search results, since the site's navigation and information architecture were incomplete or inaccessible. The homepage and about page existed but contained minimal extractable text (client-side rendered), forcing the agent to rely on search results and prior knowledge to answer what the platform does, who it serves, how it's priced, and how it differs from competitors. The site lacked a dedicated pricing page, blog, and features section, making it difficult for an agent to understand the offering without external research.

• ›Steps [1] and [4] (homepage and about page) returned HTML that was heavily client-side rendered (Next.js), yielding minimal content in the fetched response—most information was loaded dynamically and not machine-readable in the initial fetch.
• ›Steps [10], [16], and [17] (web searches) were critical: search results contained structured snippets and links (e.g., 'CELPIP Practice Test Online | Mock Exams & AI Scoring', pricing article links) that directly answered the task. The agent found pricing, features, and competitive positioning through search results, not site navigation.
• ›Steps [12] and [13] (the pricing/cost article and exams page) were fetched via search results rather than discovered through site navigation—indicating the site doesn't surface this information clearly through primary navigation.
• ›Multiple standard routes returned 404s ([3] /pricing, [5] /features, [7] /plans, [8] /blog), forcing the agent to search rather than navigate. This forced reliance on external search signals rather than on-site discoverability.
• ›The agent cited sources from both the site (main URL, about page, reviews page, cost article) and external sites (Trustpilot, App Store, CLBReady), indicating the answer required off-site aggregation.
• ›The site's agent-readiness was low: no structured data (pricing table, feature list, comparison matrix) was accessible in the HTML; no llms.txt or clear content hierarchy; dynamic rendering prevented easy extraction; and key information pages were not discoverable via standard URL patterns.