The agent gathered sufficient information to explain Qonto's core business model, pricing tiers, target audience, and differentiation, but had to rely heavily on third-party sources (search results and a competitor review blog) because qonto.eu and qonto.com served JavaScript-heavy pages that truncated in the raw response. The site itself was not machine-readable; the agent bootstrapped understanding from prior knowledge and external sources rather than direct site content.

• ›Steps [1], [2], [5], [6] fetched from qonto.eu/.com but returned only HTML scaffolding; the actual page content was rendered client-side and truncated, making them unusable as information sources. The agent correctly identified this limitation in step [7].
• ›The agent pivoted to web search (step [8]) and fetched a third-party review blog (step [9] from vivid.money) to construct its answer. Neither of these sources are Qonto's own published materials—they are competitor/review aggregator interpretations of Qonto's offering.
• ›The final response cites five external sources (OMR Reviews, Capterra, Vivid Money, GetApp, SaaSWorthy) but none are direct Qonto URLs. This indicates the agent could not surface authoritative Qonto content and had to assemble the answer from secondary sources and prior knowledge (57% of fetches came from 'previous_artifact', suggesting the LLM drew on training data).
• ›The site's structure (qonto.eu → qonto.com redirect, JavaScript-dependent rendering, 404s on /en/features and /en/business-account) created friction. The agent acknowledged missing details on per-plan feature breakdowns, API pricing, international availability, and account limits—gaps that suggest the site either doesn't expose this clearly or it's buried behind interactive UI.
• ›The agent successfully answered the 'what, who, how priced, how different' framing, but had to flag what it couldn't find, indicating partial coverage of the original task.