AI visibility: making the delivery date readable for agents
An AI assistant asked about an available item reads the product page — not your logistics system. What it finds there decides which delivery date it names.
What an AI agent reads on a product page
An agent usually receives the text of the page and its structured data. Three places count for the delivery date: the visible page text (for example “delivery time 3–5 working days”), the page’s JSON-LD with Offer and OfferShippingDetails (price, shipping cost, handling and transit time), and the field x-xl0g-delivery-date in the JSON-LD, which carries the calculated delivery date as a date. If the date is only in the text, the agent has to pick it out of a sentence. If it is machine-readable in the JSON-LD, it can take it over. If the two disagree, it is open which one the agent follows.
What the xl0g feed delivers
For each tenant, xl0g can provide a public product feed, retrievable without login and enabled on request: product data per item following Schema.org, plus the delivery date actually calculated for each size class instead of a blanket statement. The feed follows an xl0g-specific schema (v0). It is not conformant with the Universal Commerce Protocol (UCP) or the Agentic Commerce Protocol (ACP): both presuppose a shopping cart and checkout that xl0g does not operate. xl0g is not a merchant and supplies one field, the delivery date; the purchase happens in the shop. Which protocols xl0g speaks and which it does not is set out in agentic-commerce.md. A demonstration feed with invented items is available at https://xl0g.com/stage/api/ucp-feed (the path name is historical).
How the check in xl0g works
In the xl0g dashboard, two views answer two different questions. “What needs doing?” checks the product feed and the master data once a day, and does so in full: every item is looked at. For each finding the view names the item, the address and a one-sentence piece of evidence; a task counts as done when the next check no longer finds it.
“What does an agent see?” by contrast reads the real product pages, as a sample: up to 60 pages a day, the ones not read for longest first. For each page it shows whether delivery date, stock, material, EAN and dimensions are correct, missing or contradict each other in the page text or in the JSON-LD. Because one is a full count and the other a sample, the numbers in the two views do not match. The sample shows whether the pages deliver what the feed promises; it does not replace checking every page.
The first step
Open one of your product pages and look only at the source: is the delivery date in the JSON-LD, and does it match the page text? If not, that is the first task.
xl0g — Logistics without gaps. Knowledge over guesswork.