The agent assembled a comprehensive explanation of Neptune's business model, pricing, and differentiation by combining direct site fetches (homepage, service pages) with indirect discovery through the site's own comparison blog posts surfaced via web search. The site itself did not expose full page content in fetchable form—metadata and schema markup dominated the raw responses—forcing the agent to rely heavily on prior knowledge and the blog comparison articles (which were Neptune's own marketing content) to construct the final answer. The task was satisfied despite poor machine-readability of the core pages.

• ›Steps [1, 4, 5, 7, 8] returned HTML with mostly boilerplate, GTM/analytics code, and truncated content; the agent could not extract substantive product details from direct page fetches and instead relied on prior knowledge to fill the gap.
• ›Steps [10, 11] returned search results pointing to Neptune's own blog comparison articles (e.g., 'Neptune vs HelloPrenup'), which the agent used as the primary sources for pricing, positioning, and differentiators—the site's own marketing content became the most reliable information layer.
• ›The agent's final response cites five blog URLs and two main pages as sources, but only the blog posts (found via search, not direct navigation) contained the specific claims (e.g., '$5,000 flat fee', '95% confidence metric', AI concierge details) that appear in the answer; the direct page fetches did not yield this information in a structured, parseable format.
• ›The site does not publish a dedicated /pricing page (returned 404 in step [2]), forcing discovery of pricing via blog comparisons and prior knowledge rather than a single authoritative source.
• ›Key information gaps acknowledged by the agent (state availability, exact AI mechanics, service scope) reflect that even the blog content and homepage do not fully expose these details, indicating the site prioritizes marketing narrative over comprehensive, machine-readable specification.