The agent successfully assembled a comprehensive explanation of Railway by combining sparse content from Railway's own site (fetches [0–4]) with third-party comparison articles ([5], [9], [11]) sourced via web search. Railway.app's marketing pages are React SPAs with minimal readable HTML, forcing the agent to rely heavily on external sources and prior knowledge to answer pricing, positioning, and differentiation questions. The agent delivered a complete answer but had to work around the site's limited agent-readiness.

• ›Railway's own pages ([0], [1], [2], [4]) returned minified React HTML with little semantic content; meta descriptions and titles were the only machine-readable signals. The agent could not extract structured pricing tiers, feature lists, or competitive positioning directly from railway.app.
• ›The agent sourced most substantive answers from third-party comparison articles ([5], [9], [11]), which contained explicit pricing tables, feature matrices, and Render/Vercel/Fly.io comparisons—information not published on Railway's site itself.
• ›Pricing details (Hobby $5/month, Pro $20/month, Enterprise custom) came from SaaS pricing aggregators cited in search results, not from railway.app/pricing; the site's own pricing page was not content-bearing enough for the agent to parse.
• ›The agent explicitly called out that Railway's docs ([3]) were a dead-end for feature discovery and that the main site lacks comparison tables—these gaps forced reliance on external sources and prior knowledge (57% of fetches sourced from prior knowledge per metadata).
• ›The agent cited 7 external sources in its final response but only 2 Railway URLs (railway.com/features and docs.railway.com/databases), indicating that Railway's owned content alone was insufficient to satisfy the task.