The agent assembled a comprehensive understanding of ProStruct Engineering by fetching four content-bearing pages (services, cost/pricing, and about pages) discovered via web search, since direct navigation to standard URL patterns failed. The site is poorly discoverable via conventional navigation—key information is nested under non-intuitive paths and requires search or prior knowledge to locate—but once found, the fetched pages did contain substantive content about what they do, who they serve, and their differentiation. However, the agent explicitly called out a critical gap: specific pricing amounts are not publicly listed, forcing potential clients to request quotes.

• ›Steps [12], [13], [16], [19] returned functional service and about pages with details on structural engineering offerings (new homes, additions, ADUs, permits), multi-state coverage, and competitive claims (2–3 week turnaround, 100% permit approval). These were the only successful content fetches and provided ~80% of the final answer.
• ›The homepage and standard navigation paths (/about, /pricing, /services) returned 404s; the agent had to rely on web search results to discover the correct URL structure (e.g., /structural-engineers/services/, /california-structural-engineers/services/cost/), indicating poor information architecture or site reorganization.
• ›Pricing information is intentionally gated: the cost page exists and was fetched, but contains no specific dollar amounts—only a call-to-action to request a quote. This creates friction for discovery-evaluation tasks and was transparently flagged by the agent as a significant limitation.
• ›The agent used web search (steps [8], [14], [17], [20], [22]) to map the site structure and find working URLs; 36% of source material came from search results rather than site navigation, indicating the site itself does not make its own structure transparent.
• ›Content delivery relies on Next.js with JavaScript rendering, which truncated HTML responses in the logs, but the agent was able to extract enough semantic information from the fetched pages to construct a coherent narrative.