Can an AI agent buy from your store, or from your rival's?
Visa and Mastercard now run payment rails built for AI agents that shop on a person's behalf. Google lets shoppers check out inside AI Mode and Gemini. The next "visitor" to your store may never see your homepage. It reads your product data, compares you with three rivals in a second, and buys from whichever store it trusts most.
So we asked a simple question: how many stores are actually ready for that buyer?
What we found in 351 stores
We took every store RivalSweeper scored in the last 30 days: 351 ecommerce stores, mostly fashion, beauty and jewelry.
| Platform | Stores | AI-checkout profile (UCP) | Passes conformance | Blocks AI shopping agents | Blocks AI training crawlers |
|---|---|---|---|---|---|
| Shopify | 181 | 94.5% | 93.9% | 1.7% | 3.9% |
| Salesforce Commerce Cloud | 41 | 0 of 28 reachable* | 0% | 0% | 12.2% |
| Custom or undetected platform | 102 | 7.8% | 6.9% | 0% | 8.8% |
| Other platforms | 27 | 11.1% | 0% | 0% | 18.5% |
| All stores | 351 | 51.9% | 50.4% | 0.9% | 7.4% |
*13 of the 41 Salesforce Commerce Cloud stores blocked or errored on our check, so we could not see them. Salesforce also backs a different agent-checkout standard (the Agentic Commerce Protocol used by ChatGPT), which runs through a private merchant feed and cannot be seen from outside. So read this column as "UCP specifically", not "any agent checkout".
Three things stand out:
- Agent checkout is a platform story. Almost every Shopify store publishes a UCP profile because Shopify switched it on. Outside Shopify it is rare, and the big platforms are betting on different standards. If you are not on Shopify, knowing which agents can actually buy from you, and which can buy from your rivals, is now a real question to answer.
- Brands block the right bots. Very few stores block AI shopping agents (the ones that buy for a person). Far more block AI training crawlers. Those are different visitors, and it pays to treat them differently.
- Readiness is not the same as being chosen. A profile gets an agent through the door. The product data decides whether it buys.
The fine print agents read
Among stores where we checked product pages (306 stores):
| What an agent looks for | Stores that provide it in structured data |
|---|---|
| Product data (name, price, stock) | 84.3% |
| Product identifiers (GTIN) | 29.4% |
| Ratings | 24.2% |
| Return policy | 13.4% |
| Shipping details | 9.8% |
Almost every store tells an agent what it sells. Very few tell it how returns and shipping work, which is exactly where an agent decides whether a store is safe to buy from.
We ran a real race
Numbers are one thing. So we gave a shopping task to 15 real swimwear stores in one market and checked how far an AI agent could get at each: "a black one-piece swimsuit, size M, under 400 ILS, shipped to Israel." Store names are hidden.
| Store | Agent can enter | Finds a match | Reads product data | AI checkout | Result |
|---|---|---|---|---|---|
| Store A | Yes | Yes | Yes | Yes | Buys: in stock, 349.90 ILS |
| Store B | Yes | Yes | Yes | Yes | Buys: in stock, 245 ILS |
| Store C | Yes | Yes | Yes | No | Skipped: no shipping to Israel, priced in USD |
| Store D | No | Not verified | No | Yes | Skipped: blocks ChatGPT and Google shopping agents |
| Store E | Yes | No | No | Yes | Skipped: cheapest match over budget, no product data |
| Store F | Yes | No | Yes | Yes | Skipped: site search errors, size sold out |
| Store G | Yes | No | Yes | No | Skipped: no matching product, no catalog route for agents |
| 8 more stores | Yes | No | Not checked | Mostly yes | Skipped: no match in catalog, stock, sizing or region |
Two out of fifteen stores made the sale. The other thirteen were not worse brands. They were skipped for reasons a shopper would never see: a blocked bot, a size label, a missing shipping country, a search page that errors.
What an agent checks before it buys
Being reachable is step one. An agent then has to trust what it reads. In practice it looks for:
- Access. Your site lets shopping agents in (ChatGPT-User, Claude-User, Google-Agent), not only search crawlers.
- Complete product data. Price, currency, stock, sizes, color, material, identifiers like GTIN, and reviews, in structured data a machine can parse.
- The fine print. Return policy, final-sale items, shipping limits and delivery times. Clear limits read as trustworthy. Missing ones read as risk.
- A path to checkout. A conformant checkout profile so the agent can complete the order without a human.
Check your store in 30 seconds: Free AI readiness check. Then run it on your top rival and compare.
The other thing agents compare: price, including your rivals' discounts
An agent that compares four stores sees every rival promotion at once. A rival that is always "20% off" sets the reference price your full price is judged against.
That is why we track rival promotions every day. Patterns we see in real comp sets:
- Always-on discounters. Some rivals keep a sale banner up every single day of the month. Their "sale" is really their price.
- Event sellers. Others go quiet, then run deep, short sales. Those are the moments that move shoppers.
- Free-shipping players. Some never cut price and compete on delivery instead.
Each pattern calls for a different response. Matching an always-on discounter usually burns margin for little gain. Matching an event seller at the right moment can win the week.
Free calculator: what is your rivals' discounting worth to you?
Enter your monthly revenue and gross margin, then how often your top rival runs a sale and how deep it goes. The calculator shows what matching that rival would cost you each month, and which discount pattern you are facing (discount addict, teaser, event seller or full price) with the response that usually works best.
The one assumption, how much of your sales you would discount to match, is shown and editable.
What would matching your top rival cost you?
Cost of matching, per month
2,400
Per year
28,800
Share of your gross profit
4.0%
Formula: revenue × (sale days ÷ 30) × share of sales discounted × discount depth. Same currency as your revenue.
Rare, deep. Their sales are real events that move shoppers when they happen.
What usually works: Watch for their next event and plan your response in advance: a counter-offer, or a full-price launch timed away from it.
Guessing your rival's numbers? RivalSweeper tracks their promotions every day. See their real pattern in a 15-minute walkthrough.
See my rivals' real numbersThe bottom line
The next wave of ecommerce competition happens where your shoppers cannot see it: inside an agent's comparison. The brands that win it will know two things about every rival: can an agent buy from them, and what price is the agent comparing you against?
RivalSweeper tracks both, every day, for you and your rivals. Book a 15-minute walkthrough.
Frequently asked questions
What is UCP?
The Universal Commerce Protocol is an open standard that lets AI agents discover a store's catalog and complete checkout. Google and Shopify support it for checkout inside AI Mode and Gemini.
Should I block AI bots?
Block training crawlers if you want. Shopping agents (ChatGPT-User, Claude-User, Google-Agent) are buyers acting for a person. Blocking them removes you from the comparison.
Where does the data come from?
RivalSweeper monitoring: 351 stores scored in the 30 days to October 11, 2026, and product-page checks on 306 stores. The race ran on October 11, 2026, read-only (no carts or orders).
RivalSweeper Briefings
Get our competitive-intel briefings
What is changing in ecommerce pricing, catalogs, and AI visibility, and what to do about it. A short read, no fluff.
Want us to run this on your store? Book 15 minutes.