The agent delivered a comprehensive explanation of SERVE/findforcesjobs.mod.gov.uk by assembling information from two accessible external sources (GOV.UK publication and Naval Families Federation article) after the main website was blocked by CAPTCHA. The site itself is poorly navigable for agents—it blocks automated access entirely and publishes no machine-readable product information, forcing reliance on third-party documentation and prior knowledge.

• ›Step [8] (GOV.UK publication) and step [17] (Naval Families Federation) were the only content-bearing fetches; both were secondary sources, not the primary website. The GOV.UK page provided official positioning; the NFF article provided operational details like filtering options and notification features.
• ›The primary website (findforcesjobs.mod.gov.uk) returned a 400 error on the first attempt and a CAPTCHA challenge on the second, making the live job board completely inaccessible. The agent had to reconstruct the user experience and feature set from fragmented external documentation rather than from the site's own interface or structured data.
• ›60% of fetches were sourced from web search, indicating the site does not publish discoverable documentation, API docs, or structured metadata (no schema.org, no llms.txt, no /docs). The agent assembled answers from GOV.UK (government metadata) and charity/third-party guides, not from the product itself.
• ›Critical information gaps persist despite the final answer appearing comprehensive: the agent could not verify live feature sets, could not access the application process, could not confirm current opportunity counts, and could not validate the notification system's actual implementation. The table comparing SERVE to alternatives relies on logical inference, not documented differentiation.