Aivexia Tech

AX Scanner checks whether an AI agent can actually use an automotive website. Most AI-and-search tools measure whether an assistant mentions a brand; AX Scanner measures the step after: once a buyer's agent reaches the site, can it read the stock, trust the price, find the dealer and reach the test-drive form? Paste a domain. The scanner classifies it (dealer, brand/importer or online shop), reads up to 15 pages with read-only requests while respecting robots.txt, and scores it 0-100 (grade A-F) across seven layers: Discovery, Identity, Data, Actions, Trust, Agent UX and Protocol. A Muse Spark agent then attempts a real task, such as finding the nearest dealer or a car under a budget. You can replay each step and the model's reason for it, and compare what a person sees with what an agent receives without JavaScript. Every answer is verified against the real page. For dealers, it drafts JSON-LD, llms.txt and robots.txt fixes that can only be exported after human approval. The problem comes from my 30 years in automotive, across Porsche Romania and Volkswagen Group Spain contexts: sites are built for human browsers, so an AI recommendation can end at a dead end. In the sites scanned so far, 1 of 12 real brand sites and 5 of 11 real dealers passed their agent journey.