What we built is not an AI visibility product
The more precise description is this:
An end-to-end LLM evidence supply chain.
Meridian runs structured buying sequences across ChatGPT, Gemini and Perplexity. Anchored probes, generic probes, four-turn consumer conversations, eight-turn agentic simulations. It identifies the exact turn at which a brand is displaced, which competitor was selected instead, and what category of evidence gap caused the failure.
That diagnosis feeds a synthesis layer. Meridian generates model-specific JSON, conversational repair objects, citations and reasoning fingerprints calibrated to the failure type. These are not content briefs. They are evidence atoms: structured objects built to satisfy the specific constraint check the brand failed.
Each atom is published automatically to Zenodo with an immutable timestamp, a version number, a DOI and an open access licence. The atom is then routed to AI systems through robots.txt, llms.txt and platform-specific manifests.
Then the same probes run again. Same methodology. Same timestamps. The loop closes.
The result is something that did not exist before:
A brand's AI evidence layer is no longer a static collection of web pages and marketing claims. It becomes a continuously updated, publicly citable, machine-routable body of knowledge objects.
Marketing claims can be ignored by AI systems. Citable knowledge objects with DOIs, open licences and immutable timestamps are part of the evidentiary record the model draws from.
This is not content marketing.
It is evidence manufacturing for AI systems.
And the distribution is automated: from atom synthesis through Zenodo publication, structured data endpoints, llms.txt routing, review platforms, marketplace feeds and retail data APIs. Each update propagates across the evidence layer without manual intervention.
The category we are building did not have a name six months ago. We are calling it Agentic Brand Control. At its core is LLM evidence supply chain management: the discipline of building, publishing and routing machine-readable evidence objects so that AI systems select your brand, not just surface it.