The agent successfully assembled a comprehensive overview of DishwasherHero+ by combining metadata from direct page fetches, web search results, and prior knowledge—working around a React-based site that doesn't expose full content in plain HTML. It correctly identified the company's core offering (on-demand W-2 kitchen staffing), pricing tiers, target audience, and competitive positioning, but had to resort to searches and infer missing details like specific city coverage and concrete pricing examples because the site's interactive architecture hides detail pages from static fetching.

• ›Page titles and meta descriptions (steps 1–7, 20) were the only reliable content source from direct fetches; all full page content remained inaccessible due to React rendering. The agent extracted 'same-day BOH staffing in 3 steps,' 'transparent rates,' and 'W-2 model' positioning from titles alone.
• ›Web searches (steps 9, 10, 16, 17) returned higher-signal results than direct fetches—search snippets and linked pages (Instawork alternatives, G2, DishwasherHero's own blog comparison posts) provided the pricing percentages (15%/9%/7-8%), worker model details (W-2 vs 1099), and competitor context that the homepage and pricing page failed to surface as plain text.
• ›The site appears to publish comparative and detail content (alternatives page, W-2 justification page, blog posts on Qwick/Instawork comparisons) but does not make it discoverable or machine-readable via the HTML layer; the agent had to search for it or guess the URL structure. City list and concrete cost examples remain unfound despite multiple attempts.
• ›Prior knowledge and search snippets collectively filled ~81% of the answer content, while direct site fetches contributed only metadata/structure. This indicates low agent-readiness: DishwasherHero+ lacks structured data, a text-accessible FAQ, or a pricing table that a crawler can consume without JavaScript.