The agent worked around a JavaScript-heavy website to assemble a usable explanation of Vite Ma Pompe by combining direct site fetches with web search and external sources. It successfully identified the core function (free fuel price comparison for France), pricing model (free), and key competitors, but could not fully explore the site's feature set or business sustainability because the website did not serve readable content in its HTML responses and relied heavily on client-side rendering.

• ›Steps [13], [19], and [29] were the only content-bearing fetches—they yielded server-rendered HTML from data.gouv.fr, the official government fuel price site, and Indice Pompe competitor. The agent's own direct fetches to vitemapompe.fr ([1], [3], [8], [9]) returned truncated Next.js boilerplate with no visible text content, forcing the agent to rely on external sources for answers.
• ›The agent had to guess or search for core information: what Vite Ma Pompe does, its free pricing, its creator (Louis Demignot), and its iOS app were discovered through web search and data.gouv.fr metadata, not from the primary website. The final response cites external URLs as sources, indicating the site itself did not surface this information in an agent-readable form.
• ›The site's JavaScript-first architecture and lack of semantic HTML, schema markup, or text content in the initial response made it opaque to non-browser agents. Features like the 'garage' system, update frequency, business model, and App Store presence remained explicitly unclear even after multiple search attempts, showing the site does not publish structured or detailed documentation.