The agent successfully assembled a comprehensive overview of The Level by combining direct fetches of key pages (homepage, about, drug information, clinics) with web search results that surfaced contextual information from the NZ Drug Foundation. The site's information architecture was moderately navigable — core pages were reachable and content-bearing, but the agent had to rely heavily on search discovery and external sources to answer questions about pricing, positioning, and how it differs from alternatives, suggesting limited explicit competitive framing on the site itself.

• ›Steps [1, 3, 9, 10, 15, 16] returned substantive HTML with page titles and metadata (e.g., 'Straight up drug info', 'Drug information', 'Drug Checking Clinics') that anchored the agent's understanding of what The Level offers, but the actual page body content was truncated in the responses, forcing reliance on metadata and inferred structure.
• ›Steps [12, 13] (web searches) were critical to answering 'how it's priced' and 'how it's different' — the agent found no dedicated pricing or comparison pages on the site itself and had to bootstrap context from the NZ Drug Foundation's own announcements and press materials. The site does not explicitly position itself against alternatives or articulate its unique value prop in a machine-readable way.
• ›The agent discovered the drug checking clinics feature ([16]) and resource ordering ([12] search snippet) through search results and prior knowledge rather than clear site navigation, indicating these services exist but are not prominently surfaced in the main information architecture. Two attempted URLs ([6, 7] /about, /contact) returned 404s, revealing shallow IA.
• ›The site is citation-friendly — the agent was able to cite six specific URLs from its fetches and searches — but lacks structured metadata (schema.org, llms.txt, or FAQ markup) that would allow a machine to extract 'what we do', 'pricing', 'alternatives' directly without assembly work.