The agent successfully gathered the core information needed to explain Internacious to someone else by relying primarily on a single well-structured resource: the site's llms.txt file. While the agent fetched multiple pages (homepage, services, about, contact), those pages returned only boilerplate Next.js HTML with no rendered content; the llms.txt file at step [13] was the only content-bearing source that directly answered all four questions (what they do, who it's for, pricing, differentiation). The site is moderately agent-hostile: it publishes useful structured information via llms.txt but does not provide a traditional pricing page, SLAs, case studies, or directly comparable competitor positioning, forcing the agent to work from metadata and inference.

• ›Step [13] (llms.txt fetch) was the sole content-bearing resource that answered the task completely. It contained: service list, target market (Sydney/Canberra, 5–100 employees), fixed pricing ($150–225/user/month), founding year (2000), and key differentiators (local support, no lock-in, Essential Eight alignment). All other fetched pages returned JavaScript-heavy Next.js boilerplate HTML with no readable content.
• ›The site deliberately does not publish a dedicated /pricing page (404 at step [4]); pricing information is only accessible via llms.txt metadata, not through standard web navigation. This signals that pricing is intentionally gated behind contact/conversation rather than self-service discovery.
• ›The agent discovered llms.txt through prior knowledge / direct guess (not by being linked from any page), indicating the site did not surface this resource organically. The sitemap.xml (step [6]) listed pages but not the llms.txt file, so the agent had to know to look for it.
• ›The agent noted two small inconsistencies (5–75 vs. 5–100 employee range) and could not retrieve SLAs, case studies, or testimonials, indicating incomplete content depth even in the best available source.
• ›The site is a Next.js SPA that renders content client-side; raw HTML fetches showed only framework boilerplate. This made traditional page-by-page navigation ineffective for automated agents. llms.txt was the only machine-readable fallback.