Agent assembled a partial picture of Alixon by combining sitemap structure, page titles, and one web-search summary that included body-text paraphrasing; the site's heavy reliance on client-side rendering blocked direct fetches of content-bearing pages, forcing the agent to infer offerings from titles and rely on search results to answer the core task.

• ›Sitemap.xml (step [4]) was the only content-bearing resource the agent fetched directly from the site; it revealed 32 URLs (services, solutions, case studies, blog) but no substantive content. Nearly every fetch of those pages (steps [1], [6], [7], [8], [10]) returned title tags only, indicating JavaScript-rendered or dynamically-loaded body content that the agent's fetch tool could not execute.
• ›The web-search result snippet (step [9]) was the only place the agent found actual service descriptions and the founder's background—"connect AI agents to a client's existing tools to cut repetitive tasks, missed follow-ups, and processing delays" and "5 ans d'expérience en PME." This content came from Alixon's own pages but was mediated by a search engine, not fetched directly.
• ›Pricing is definitively absent from the site's information architecture: /tarifs and /services both returned 404, and the ROI calculator URL in the sitemap could not be read. The agent correctly identified this gap rather than speculating.
• ›The agent faced a multi-domain landscape (alixon.fr, www.alixon.fr, alixon.vercel.app) without a canonical declaration and could not determine which was primary; the sitemap referenced www.alixon.fr, but the agent had no way to verify if content diverged across mirrors.
• ›The agent's final answer relied heavily on inference from page titles, sitemap structure, and a single search snippet, all of which are fragile sources for a discovery task. No case study, testimonial, or product detail page was actually readable, limiting confidence in claims about what Alixon builds.