On 15 October, Google switched on Universal Cart for shoppers in the US. You add products from any store to one cart while you search or chat with Gemini, and Google keeps watch for you. It tells you when a price drops or when something comes back in stock, and at stores that support it, you can check out without leaving Google. Walmart, Wayfair and Ulta Beauty were in on day one.
Now look at it from the store's side. To watch your price and your stock, Google has to be able to read them. If it can, your product sits in the cart, tracked and a tap from checkout. If it cannot, there is nothing to watch.
So we measured how many stores are in that position, and the answer is further down. First, the bigger picture. Universal Cart is one agent, and your customers will use many: ChatGPT, Perplexity, Copilot and the assistants arriving inside the apps they already use. Each reads your store its own way, and each decides for itself whether your product makes the shortlist.
Omni-agent readiness is how ready your store is for all of them.
What is omni-agent readiness?
Omni-agent readiness is a measure of how well AI agents can find, read, trust and buy your products, whichever agent your customer happens to use.
It is the measurable side of omni-agent commerce: the discipline of being legible, comparable and buyable by every AI agent a customer might delegate to, regardless of who built the agent.
A store can look perfect to a person and still be hard for an agent to use, because agents do not look at your pages the way shoppers do. A shopper sees the photos, the price and the "in stock" badge. An agent reads the data underneath, and makes its decision on that.
Is my store ready for AI shopping agents?
We look at readiness in five areas, each one a question an agent effectively asks of your store.
- Can it find your products? Whether agents are allowed in, and whether your site leads them to your product pages.
- Can it read the basics? Whether each product page states its name, price and stock in a form a machine can read, not just in text a person sees.
- Is the data complete and consistent? Whether the details are all there, and whether what your data says matches what the page says.
- Can it understand and trust it? Whether descriptions answer a shopper's real questions, and whether signals like ratings back the product up.
- Can it buy? Whether the product is actually for sale at a real price, so an agent could take a shopper from question to checkout.
We bring these together into one number from 0 to 100: the Omni-Agent Readiness Score. Some of these areas can be checked from outside your store, and some need your own product data, so a score from a public scan and a score from a connected store can differ. We always say which one you are looking at.
What readiness is not
Readiness is not a promise of recommendation. Readable means the agent can see you. No tool can guarantee it will recommend you.
It also should not be a guess. If a check cannot see part of your store, it should say so rather than score it. Ours only scores what it actually observed, so a store is never put in a lower band because we could not look, and never gets credit for what we did not see.
What we found when we measured the basics
To see how common the gaps are, we ran our scan engine across a random third of every US and UK online store that is not on Shopify. That is about 94,000 stores, drawn from a list of 288,171.
We estimate that about 7,600 of those 288,171 stores are missing price or stock from their structured product data on half or more of the product pages we sampled. The 95% interval runs from 5,000 to 10,600. About 2,800 of them are in the $1M to $20M revenue band. Stores we could not read are not counted.
That is only the most basic test: whether price and stock are there at all. It is not a readiness score for those stores, and it says nothing either way about Shopify stores, which we have not yet measured.
Why does Google reject products for missing price?
When price or stock is missing from that data, the product is hard for a machine to trust. Google checks this data when it reviews products for Shopping. An AI shopping agent comparing options cannot confirm a price or availability it cannot read, so it moves on to a store where it can.
Because these stores are missing it on half or more of the pages we read, it is rarely one broken page. It is usually a large share of the catalogue.
Our first pilot store had 588 products. Google had rejected all 588, and nobody in the business knew.
How we measured it
The order of stores was randomised and fixed in writing before we saw any result. For each store, our scan engine reads its sitemaps and up to 25 product pages spread across the catalogue, and checks whether price, stock and a product ID are present in the machine-readable data.
A flag from the engine is not the same as a confirmed problem, so we checked a random sample of findings page by page against what a shopper actually sees. We only counted stores selling goods online with a visible price. Quote-only sites, services and out-of-stock pages do not count. The estimate above is after that correction.
Check your own store in a minute
Paste one of your product URLs into Google's Rich Results Test. If price and availability do not appear under Product, an AI agent cannot read them either.
That answers one of the five questions. To see how your store does across all five, book a call and we will walk you through what AI agents can read about your products today.
You're one call away from making sure AI knows you exist.
We'll show you what AI shoppers can read about your products today, and what they can't.
Book a call