The agent successfully compiled a comprehensive explanation of Twenty.com by assembling information from multiple sources, but had to heavily rely on web search and external documentation because the main website's pages are JavaScript-rendered and returned truncated, unreadable HTML. The site itself is poorly navigable for machine agents—most navigation paths returned 404s, and the key pages (homepage, pricing, product) did not yield useful text content in the fetches, forcing the agent to source 72% of its answer from web search results and prior knowledge rather than direct site content.

• ›Steps [1], [3], [11], [12], [18] fetched the main site URLs but all returned heavily truncated HTML with preloaded fonts and Next.js bundled JavaScript—no extractable content about what Twenty does, pricing, or positioning. The agent explicitly cited these pages as sources but likely drew conclusions from search snippet metadata rather than the actual response bodies.
• ›Steps [4], [5], [8], [9] tested common navigation paths (/features, /docs, /about, /open-source) and all returned 404s, indicating a shallow site structure or poor crawlability. The site does not surface its own documentation through main-domain pages.
• ›The agent pivoted to web search (steps [7], [15], [16], [21], [22], [24], [25]) and found third-party reviews, GitHub repo summaries, and external CRM comparison sites more useful than the primary domain itself. Step [13] successfully fetched the external documentation site (docs.twenty.com) which returned readable content, but this is hosted separately and not discoverable from the main site.
• ›The agent cited seven external sources (pricing page, docs.twenty.com, product page, comparison pages, GitHub, opensourcealternatives.to, and DEV Community) in its final response but most of these were identified through search results, not by navigating the site itself. The main twenty.com URLs were cited but appear to have been treated as confirmatory rather than content-bearing.
• ›The agent's final answer is comprehensive and accurate but is a synthesis of prior knowledge (~36%), web search findings (~36%), and fragments from earlier fetches (~27%). This distribution reflects that Twenty.com's website is not the primary source of truth for an agent trying to understand the product—external resources and documentation sites had to fill the gap.