The agent assembled a comprehensive explanation of WhatsApp for Business by combining direct site fetches with web search results, supplemented by prior knowledge. The official site itself was poorly navigable—initial direct guesses for /pricing, /features, and /solutions all returned 404s, and the HTML responses were truncated, providing no usable content. The agent had to rely heavily on web search to discover the correct URLs (/products/platform-pricing, /products/business-platform, /resources/faq) and fill gaps about pricing models, the two-product distinction, and competitive positioning that the main site did not prominently expose.

• ›Direct URL guessing failed: /pricing, /features, /solutions returned 404s. The agent only successfully fetched pages after discovering them via web search (e.g., /products/platform-pricing, /resources/faq), indicating poor information architecture or discoverability.
• ›HTML responses were truncated in all fetches, rendering them unreadable for machine extraction. The agent relied on prior knowledge (36%) and web search results (36%) rather than parsing the actual site content, suggesting the site either heavily JavaScript-renders content or does not structure pricing/feature information in a crawlable form.
• ›Two-product model (Business App vs. Business Platform) was not clearly distinguished on the main site. The agent discovered this via web search queries and had to infer the distinction rather than find it explicitly laid out on the homepage or a dedicated comparison page.
• ›Pricing opacity: The site acknowledges country-based pricing but does not display actual rates or examples. Users must contact sales or use a Business Service Provider intermediary, making pricing discovery difficult for both humans and agents.
• ›Missing technical bridge: The site does not explain the requirement to use a Business Service Provider (BSP) intermediary to access the API, a critical friction point that the agent had to infer from external sources.
• ›Site heavily relies on blog and resource content (fetched from /blog, /resources/faq, /resources/resource-library) for substantive product explanation rather than dedicated product pages, scattering key information across multiple content types.