Agentic Landmark · The Brief
Briefs on the infrastructure, authorization architecture, and commerce strategy of the agentic web: what brands need so agents can find, evaluate, and transact with them.
Google has published six optional Merchant Center fields that tell its conversational surfaces what to know about a product. What each one carries, which to fill first, and the rule that keeps the feed from replacing the page: the values have to match.
In a recorded test on fictional storefronts, an AI shopping agent passed over the highest-rated shoe in five of five runs. Its specs were in an image. This Brief covers the structural floor under every discovery channel, and the free instrument that reads it.
Amazon shut Meta's Muse agent out and Shopify let it in, twelve days after quietly switching the channel on for roughly a million merchants. Both platforms made the same decision on behalf of every brand underneath them. What the default turned on, and what to do this week.
Google's holiday guidance tells retailers to optimize for people, not bots. Its other recommendations are all written for machines: richer product feeds, real-time price and inventory, agentic checkout. The contradiction resolves once you separate the surface layer from the infrastructure layer.
Fifteen minutes to a working commerce agent. The repository's own note hands authorization, business rules, and compliance back to the deployment, and that is the part that decides whether any of it should be running.
Think with Google told brands that foundational SEO remains the formula for AI search visibility. In the same article, DoorDash describes integrating 500,000 SKUs with merchant-level availability so Gemini knows what is actually in stock. The example refutes the headline.
McKinsey published two figures for brand-owned share of LLM citations eight months apart. They differ five-fold, and the firm reconciles neither. The gap is not a scandal. It is what happens when an industry builds a capability around a number nobody can measure twice.
Anthropic now marks Claude output with an embedded text watermark and C2PA metadata. The documentation is candid about what that proves, which is very little. Provenance and authorization are different records, and only one of them is being built for you.
Both agent readiness instruments produce a weighted composite, and an average is a claim about the whole. A break is a claim about one link. Those do not commute. Why the score tells you what your exposure costs, and only an ordered read can tell you where you stop.
An agent and a seller want the same things right up until the recommendation, which is the one moment that decides the sale. Whoever funds the agent decides what happens there. For the agents that earn the highest-value delegation, there is no rate card, and there was never going to be one.
