The agent successfully gathered comprehensive information about Mercury's business model, pricing, target audience, and competitive positioning by combining direct site fetches with web search and third-party reviews. The mercury.com site itself was heavily JavaScript-rendered and marketing-focused, making it difficult to extract structured product details; the agent compensated by relying on prior knowledge (57% of fetches sourced) and external review sites (NerdWallet, Brex, Rho) that provided clearer feature comparisons and pricing breakdowns than Mercury's own pages.

• ›Steps [1], [2], [5] fetched Mercury's homepage, pricing, and about pages directly but returned JavaScript-heavy HTML that the agent could not parse for detailed content—the agent explicitly noted this limitation in step [7] and pivoted to web search.
• ›Step [8] (web search) was a routing step, but its result snippets and linked URLs (especially mercury.com/pricing and mercury.com/business-banking) provided the first structured signal; the agent then used step [9] (NerdWallet review) as the primary source for detailed features, pricing tiers, FDIC limits, and competitor comparisons—information NOT clearly laid out on Mercury's own site.
• ›The agent's final answer draws heavily on prior knowledge and external reviews rather than Mercury's own published content; Mercury's site lacks a machine-readable, structured pricing table, feature comparison, or clear positioning statement against competitors, forcing the agent to synthesize from third-party sources.
• ›Multiple attempted navigation paths ([3] /solutions, [4] /features, [6] /resources) returned 404s, indicating a sparse site structure and reliance on the homepage/pricing/about to communicate value.