The agent successfully gathered enough information from two homepage fetches to explain Opendoor's core business, products, pricing approach, and key tradeoffs. The site made basic positioning and process flow discoverable, but fragmented detailed information across the homepage and a secondary product page while leaving fee amounts, eligibility criteria, and service areas unstated—requiring the agent to note what it could not find.

• ›Step [0] (homepage) and step [1] (Cash Now, More Later product page) were the only fetches performed; the agent typed the homepage URL from memory and followed one internal link. The site's IA does not surface a dedicated pricing page, how-it-works explainer, or fee disclosure—forcing the agent to assemble facts from fragments and flag gaps.
• ›The homepage contained corrupted or placeholder text (customer count with run-on digits, rate strings like '0123456789…', and 'handlesmarketingrepairs' with no spaces), which the agent correctly identified as data quality issues that would confuse users.
• ›Mortgage rate information on the homepage conflicted (30-year fixed at 7.450% vs. featured ARMs at 4.99%), and the agent traced this to fine-print requirements (20% down, 780 credit score, affiliate financing) rather than guessing; this was site-supplied detail, not model memory.
• ›The agent identified that fee percentages, formulas, and second-payment calculations are not published by the site, forcing sellers to request an offer to see actual costs—a deliberately opaque pricing model that the agent correctly attributed to the site's structure, not to missing research.
• ›The agent did not visit the help center, FAQ, or licensing pages despite links being available; its final answer was drawn entirely from the marketing homepage and the Cash Now, More Later product page, demonstrating that the site's top-level pages alone did not fully answer the task, yet the agent delivered a substantive explanation by clearly flagging what remained unknown.