The agent successfully assembled a comprehensive explanation of Swan AI by fetching the company's homepage, pricing page, and about page, then triangulating with third-party review and comparison articles. The site itself is reasonably navigable and publishes core information (positioning, pricing tiers, capabilities) in structured pages, but the agent had to rely heavily on external sources (71% web search) to fill gaps around feature details, competitive positioning, and concrete pricing examples—suggesting Swan's own site is positioning-forward rather than comprehensively self-documenting.

• ›Step [5] (homepage) and [6] (pricing page) returned Swan's official positioning ('AI GTM Engineer,' credit-based model) and pricing tiers ($80–$419/month), which form the backbone of the answer. These were directly fetchable and well-structured.
• ›Step [7] (about page) provided founder background and funding context, showing the site does publish company identity; however, the agent relied much more heavily on external reviews ([10], [11], [14]) to explain *what Swan actually does* in practical terms (lead qualification, CRM automation, outreach sequences)—suggesting Swan's own pages emphasize positioning over feature documentation.
• ›The agent explicitly cited all six content-bearing steps in its Sources section, and none were 'routing only'—each contributed substantive information to the final answer. However, 71% of fetch decisions came from web search, not site navigation, indicating the homepage and pricing page alone do not comprehensively expose feature details, integrations, or competitive context.
• ›The agent's 'Things That Were Confusing or Missing' section—credit consumption opacity, scattered plan definitions, no ROI examples, unclear learning curve—directly reflects gaps in what the official site published vs. what third-party sources had to fill in. This is a candid signal that Swan's own site prioritizes pitch over documentation.
• ›The site is agent-navigable (clean URLs, published pricing page, accessible meta descriptions) but not comprehensive; an agent must combine official pages with external review aggregators to answer the task fully. This is typical of early-stage SaaS but places friction on discovery and identity layers.