The agent successfully assembled a comprehensive explanation of Linktree by combining the site's own llms.txt file (which contained structured product definition and feature overview) with external third-party sources about pricing, alternatives, and AI features. The official website itself was poorly machine-readable—fetches [1], [2], [3], [6] returned React framework code rather than meaningful content—so the agent relied heavily on Linktree's metadata file and web search results to answer the task. The site's agent-readiness was mixed: it publishes useful machine-readable summaries (llms.txt) but obscures marketing copy and pricing details behind client-side rendering.

• ›Step [5] (llms.txt) was the only directly fetched page that returned structured, parseable content about what Linktree does, who uses it, and core features. This became the skeleton of the answer; without it, the agent would have been entirely dependent on web search.
• ›Steps [1], [2], [3], [6] fetched the actual marketing website but returned React boilerplate with minimal readable text, forcing the agent to bypass the primary domain and rely on external sources for pricing [8, 13, 16] and competitive positioning [13, 16].
• ›The agent did not extract pricing details or competitor comparisons from Linktree's own pages—those came entirely from third-party review sites cited in steps [8] and [13]. Linktree's official pricing page [2] was fetched but not cited, suggesting it was not extractable or was less useful than the external sources.
• ›The agent discovered llms.txt through prior knowledge (direct guess), not by following links from the homepage, indicating this artifact exists but is not discoverable through normal navigation—it required explicit knowledge of the pattern.
• ›Sales commission details, the main pricing differentiator between plans, were sourced from external guides [8, 13, 16], not from Linktree's own site, and the agent correctly flagged this as confusing and buried in the final response.