The agent successfully assembled a comprehensive explanation of Ramp's business model, pricing, and competitive positioning, but had to rely heavily on web search and prior knowledge because Ramp's own website served machine-optimized markdown and truncated HTML responses that did not contain usable product information. The agent pieced together the answer from third-party review sites, comparison articles, and blog posts rather than from Ramp's primary content.

• ›Step [1] returned a 'Machine Version' homepage in markdown format with only a promotional offer and quick links—no substantive product description. Steps [2], [4], [5] returned raw HTML with truncated responses (4400+ chars cut off), making them unusable without rendering. The agent could not extract pricing or feature details from Ramp's own pages.
• ›All substantive content came from external sources: Capterra (step [6]), Precoro and Ramp's own blog comparison (step [7]), third-party guides and reviews (steps [10], [12], [13]). Step [11] (Ramp's Expensify vs. BILL vs. Ramp blog post) was the only Ramp-authored page successfully cited, but its response was also truncated HTML.
• ›The site lacks agent-ready structured data, a clear product overview page, and transparent feature-by-tier documentation. The agent had to search externally to answer basic questions about what Ramp does, who it targets, and how it's priced—information that should be immediately available on the homepage or dedicated product pages.