Agentic commerce
Unstated means eliminated.
When a buyer gives an assistant a constraint, the assistant filters the catalogue before it ranks anything. An item whose price it cannot read is not ranked lower. It is dropped, before a human sees the result. Antahe checks the fields that decide that, on the catalogue you actually publish.
This is not a bet on the future. It is a gap in the present.
Google announced the Universal Commerce Protocol on 11 January 2026 and has been onboarding merchants through Merchant Center since March 2026. It standardises discovery, not just checkout, which is the part that decides whether you are considered. The same connection that tells us how complete your catalogue is tells us where you stand on it.
Source: Google, Universal Commerce Protocol, announced at NRF on 11 January 2026. Merchant Center onboarding since March 2026.
On the sentence at the top of this page: no vendor publishes the rule, and we are not claiming one does. A filter removes what does not match before a ranker orders what is left, which is what makes an unstated field an exclusion rather than a penalty. What is ours is the list below of which fields behave that way, and it is the list our checker uses.
Seven fields decide whether you are considered at all
These are the ones an agent filters on. A missing value here is not a weaker listing, it is no listing: the item leaves the candidate set before ranking starts. Read from the same table the checker uses, so this page cannot describe a check we do not run.
| Field | What it costs you to leave blank |
|---|---|
| price | An agent told to stay under a budget cannot include an item with no price. It is dropped, not ranked lower. |
| currency | A price without a currency is not comparable, so it is treated as absent. |
| availability | Agents exclude anything they cannot confirm is in stock rather than risk a failed order. |
| shipping | Delivery is a hard constraint in most agent shopping flows, and an unknown is a no. |
| returnPolicy | Return terms are a common filter and a common reason to be excluded silently. |
| condition | New versus refurbished is a filter in most catalogues. |
| link | An item an agent cannot reach cannot be bought. |
Five more decide where you place once you are in
These cost ranking rather than eligibility, which is a real difference and a different morning's work. Fixing a filterable field buys you candidate slots. Fixing one of these improves your position in the slots you already have.
| Field | What it costs |
|---|---|
| gtin | Without a standard identifier an agent cannot match your item to the same product elsewhere. |
| brand | Brand-level filtering and disambiguation both need this. |
| title | The primary matching signal for intent. |
| description | Where constraints a shopper might filter on usually live in prose. |
| imageLink | Absent images reduce selection in surfaces that show them. |
If you sell a service, the same rule holds with different fields
A consultancy has no stock level and an insurance policy has no barcode, so we do not ask for them. What an assistant needs is the same in shape: what it is, who it is for, what it costs or how pricing works, where the engagement happens, and how to start one. Antahe decides which set applies from what you sell, and freezes that choice on your first audit so your results stay comparable run to run.
The sentence no competitor is saying
You are enrolled in UCP. You are not ready for it. Six of the fields an agent filters on are missing, so you are dropped from the candidate set before a human sees the result.
Enrolment is not optimisation, and the gap between the two is the finding. Being onboarded to a protocol says a surface can reach your catalogue. It says nothing about whether that catalogue answers the question a buyer asked.
We can measure the readiness half today. Enrolment state is not exposed by the Content API, so where we cannot read it we say so rather than guessing.
Questions we get asked
What is agentic commerce?
Agentic commerce is a purchase where software does the shortlisting. A buyer tells an assistant what they want, with constraints like a budget or a delivery date, and the assistant filters a catalogue before it ranks anything. Google's Universal Commerce Protocol, announced on 11 January 2026 and onboarding through Merchant Center since March 2026, is the standard that matters, because it covers discovery rather than only checkout.
How do I get ready for agentic commerce?
Publish every field an agent filters on, because an unstated field is an exclusion rather than a lower rank. An assistant told to stay under a budget cannot consider an item whose price it does not know, so it drops it before a human sees the result. Price, currency, availability, shipping, returns, condition and a working link are the seven that decide the candidate set; title, description, brand, images and a standard identifier decide the ranking after that.
Is being ready for AI different from SEO?
It overlaps and it is not the same. Crawlability, structured data and page speed help both, and a page an AI crawler cannot read is usually one a search crawler struggles with too. What is different is that assistants do not execute JavaScript: a page that builds its content in the browser is a full page to a person and close to empty to a crawler, and that gap is invisible in analytics because analytics runs in the browser where everything already worked.
See where you stand
The free check reads your site as an AI crawler does and tells you what it could not read. The paid audit adds the catalogue: which of your items an agent could shortlist, and which field is dropping the rest.
The check takes a domain and no account. It reads your site as a crawler does and shows you the result before it asks for anything.