The agent gathered a partial understanding of Woven by fetching the homepage and pricing page, then relied heavily on prior knowledge and web search results to fill gaps left by a minimal website. The site itself surfaced only basic pricing information; the agent had to assemble claims about features, positioning, and differentiation from external sources and inference rather than from Woven's own documentation.

• ›Steps [0] and [1] (homepage and /pricing) were the only successful direct fetches from woven.video and contained the actual pricing ($8.25/month annual, 3-day free trial) and high-level positioning as a macOS AI video editor. However, HTML responses were truncated and the agent did not extract specific feature details from them.
• ›The agent attempted to discover deeper content via direct guesses (/features, /about, /faq, /download, /blog) and all returned 404s, indicating a minimal site structure. Attempts to fetch external help resources (docs.woven.video, help.woven.video, support.woven.com) all failed, leaving no discoverable documentation.
• ›The agent compensated by relying on prior knowledge (43% of fetch sources) and web search (29%). Search results consistently returned the homepage but never returned Woven-specific comparison articles, reviews, or feature documentation—suggesting Woven has minimal SEO presence or online discussion. The agent inferred most competitive positioning ('conversation-based editing as primary interface', 'native macOS', 'bring your own keys') from domain knowledge and general knowledge of AI video editors rather than from Woven's own content.
• ›The site exposed minimal information in a machine-readable form: pricing page loaded but its HTML was not parsed into structured data by the agent. The agent's final answer includes several explicit caveats about what it 'couldn't fully determine'—specific competitive advantages, output quality, documentation, maturity, and feature limits—all due to inaccessible or absent documentation on the site itself.