A multi-location car-rental brand had thirteen separate sites that AI engines saw as thirteen strangers, with no way to know whether AI ever sent a real customer. We fixed both: made the network legible as one linked entity, graded every site, and turned on attribution.
Two problems compound in a location network. First, fragmentation: thirteen sites with no shared identity read to an AI engine as thirteen unrelated small players, none strong enough to be the answer. Second — the one every tool ignores — no proof: even where AI mentioned a location, there was no way to know whether a single real customer ever walked in from that answer. Citations and “AI reach” look nice on a dashboard; they don't fill a rental slot.
Organization + subOrganization schema with bidirectional @id/sameAs, so all thirteen sites resolve as one linked brand — the network gets the authority its size deserves.
Per-site schema, llms.txt, MCP manifest and AI-friendly robots — the same deploy-ready package applied across every location, each graded on the open-book NVS to Grade B.
A lightweight beacon on every page that fires only for a real human arriving from an AI engine — turning “are we cited?” into “did AI send us a customer?”.
The beacon separates AI-bot crawls (“AI can see us”) from real human arrivals (“AI recommended us”), so the client reads two honest numbers, not one blurred one.
A ping the first time an AI answer-engine indexes a site, and again when a real person arrives from AI — the moment the machine layer starts paying off.
Thirteen deploy-ready packages on the client's own domains. The AI identity and the measurement stay theirs.
Monitoring platforms tell you how often an engine mentions you. That's the start, not the finish. Our beacon measures the end of the funnel — a real person who arrived because an AI recommended you — per location, live. It's the difference between “we're visible” and “AI is bringing us customers,” and it's the number a network owner actually renews on.