The agent successfully built a comprehensive explanation of a7.de's business model, target audience, pricing, and competitive positioning by combining the homepage (step [1]) with the machine-readable llms.txt file (step [2]). The site was moderately navigable: it published a structured, LLM-friendly summary in llms.txt that contained most answer material, but pricing information was fragmented across the site and the agent had to rely heavily on that single artifact rather than discoverable service pages.

• ›Step [2] (llms.txt) was the critical source—it contained a dense, well-organized summary of the company's services, target audience (50–5,000 employee B2B firms in DACH), pricing tiers, certifications, and founding narrative. This single artifact answered ~80% of the task without the agent needing to parse individual service pages.
• ›Step [1] (homepage) was cited as a source but appears to have been minimal—most of the final answer's specifics (pricing, certifications, NPS score, employee count) came directly from llms.txt, not from HTML parsing of the public pages.
• ›Steps [3] and [4] (pricing page attempt and penetration test service page) returned 404 or truncated HTML. The agent was blocked from scraping service detail pages, forcing reliance on the llms.txt artifact.
• ›The site explicitly publishes instructions in llms.txt for how LLMs should interpret and cite AWARE7 content, signaling intentional agent-readiness—but this also suggests the site owners expect agents to rely on that file rather than traditional HTML navigation.
• ›Pricing fragmentation was a real friction point: no dedicated `/preise` page exists; the agent noted that several services (phishing, ISO consulting, dark web monitoring) require manual quotes and pricing is 'scattered across service pages and the llms.txt file'—but in practice, llms.txt was the only accessible source that contained concrete numbers.
• ›The agent correctly identified what was confusing or missing: terminology choices ('Datenleck-Monitoring'), lack of public pricing for some services, and unclear enterprise-vs-SMB positioning—all gaps that existed in the retrieved content, not due to agent error.