The agent successfully synthesized information about course.accompany.tw's purpose, audience, pricing, and differentiation by combining web search results with a markdown course page fetch. The site's JavaScript-heavy structure made direct HTML inspection unproductive, forcing reliance on search indexing and a markdown fallback route; pricing and competitive positioning had to be assembled from fragments rather than published in a clear, structured format.

• ›Step [5] (web search) surfaced the course catalog and instructor profile URLs, establishing the site's identity as an OT-led parenting academy. The search results themselves contained minimal detail but acted as a discovery layer pointing to specific courses.
• ›Step [10] (markdown fallback) was the only content-bearing fetch that returned readable, structured information—course title, description, tags, and course type (paid). This was the primary source for understanding what the platform teaches; the HTML pages (steps [1], [3], [7], [8]) returned only script and CSS bundles with no extractable content.
• ›Pricing information (NT$3,550–NT$4,550) and tier details were embedded in the markdown frontmatter and course description but required the agent to parse markdown instead of fetching a dedicated pricing API or landing page. No master course catalog or comparative pricing matrix was discoverable.
• ›The agent had to reconstruct competitive differentiation (OT credibility, PEO framework, multi-platform ecosystem, podcast ranking) by synthesizing scattered references from the course description and prior knowledge about the founder—the site does not publish an explicit 'why us' or competitor comparison section.
• ›The site's primary content is rendered client-side (Next.js), making it opaque to direct HTTP fetching. The presence of a `.md` route suggests the platform indexes content in markdown but does not surface it as a public content API or structured data format (no JSON-LD, OpenGraph, or schema.org markup visible in truncated HTML).