The agent successfully compiled a comprehensive explanation of Vercel's purpose, target audience, pricing tiers, and competitive positioning, but had to rely almost entirely on web search and prior knowledge rather than Vercel's own website. The direct fetches of vercel.com (steps 1–5) returned heavily JavaScript-minified HTML that was not machine-readable; the agent pivoted to external sources (three web searches) to construct the answer, which were then cited in the final response.

• ›Steps 1–5 (direct site fetches) returned only minified, JavaScript-heavy HTML with no extractable content; the agent correctly identified this as unusable and abandoned the direct crawl approach.
• ›All substantive information came from web search results (steps 6, 7, 8) linking to third-party review/comparison sites (Gartner, GetApp, SkillScouter, Kuberns, CostBench, Index.dev, Encore, etc.), not from Vercel's own published pages. The agent did not retrieve readable content from vercel.com/pricing, vercel.com/features, or vercel.com/docs despite attempting them.
• ›The final response explicitly cites 11 third-party sources and zero primary Vercel documentation, indicating Vercel's own website did not surface answers in an agent-readable format (no structured data, no plaintext alternatives, no static fallbacks).
• ›The agent reconstructed pricing tiers, feature comparisons, and competitive positioning from aggregated external reviews rather than canonical Vercel sources, introducing reliance on third-party interpretation and potential staleness.
• ›The agent noted confusions and gaps it encountered (pricing calculation complexity, hidden costs, unclear feature matrices, lack of cost calculator) — these reflect gaps in what Vercel's site itself exposed, not what the agent failed to find.