Agentic commerce readiness
Can AI shopping agents actually find your store?
Assistants are starting to shop on people's behalf. They do not browse your storefront the way a shopper does: they look for a protocol profile at a fixed path on your domain, read your structured data, and move on if either is missing. This checks whether yours is there.
Free, no account. Reads only public pages, the same way an agent would.
Validated against the published UCP specification, currently 2026-08-25. We check what the spec says, not what a vendor blog says it says.
The four checks
Four things decide whether an agent can buy from you
- 01
Agent discovery
Whether a UCP profile exists at /.well-known/ucp, parses, declares services and capabilities, and is served the way the spec demands: HTTPS, no redirect, public cache header.
- 02
AI crawler access
Whether robots.txt quietly blocks the crawlers behind ChatGPT, Gemini, Claude, Perplexity or Alexa. Blocking these removes a store from those surfaces outright.
- 03
Product data agents can read
Whether product pages carry schema.org Product markup with a complete offer, and whether products carry the identifiers agents use to match and compare them.
- 04
Catalogue reachability
Whether a sitemap and a machine-readable product endpoint exist, so an agent can learn the whole catalogue rather than the pages it stumbled on.
Reference
What we know about this, written down
The protocol is new and most of what is written about it is second-hand. These pages quote the specification directly and show the commands to check your own store. They are useful whether or not you buy anything, and that is deliberate.
- /.well-known/ucp explainedWhat the file is, the five hosting rules the spec sets, and how to serve it on WooCommerce, Magento, PrestaShop, BigCommerce or a headless stack.
- Why redirects break discoveryThe spec forbids a 3xx on the profile endpoint. That one line rules out the workaround most platforms leave you, including apex-to-www.
- Which AI crawlers to allowThe exact user-agent tokens, what each one really gates, and which blocks remove a store from a shopping surface rather than from a training set.
- Product data agents can readThe offer fields and identifiers an agent needs before it can price, compare and propose a product, separated from the ones only Google uses.
- Our own score: 69/100We ran our own scanner on storedeploy.com and published the result, including the finding we have decided not to fix.
Monitoring
A scan tells you about today. This tells you when today stops being true.
Platforms update, themes get replaced, someone edits robots.txt. Any of those can undo the four checks without anyone noticing, because nothing visibly breaks. We run the checks every morning and write to you when the answer changes.
- The same four checks, run against your store every morning
- An email only when something actually changes: a new critical or high problem, or a drop of ten points or more
- A quiet inbox means nothing broke. We do not send a report just to look busy
- If a bot filter blocks the scan, we say so and skip the comparison instead of inventing a regression
- No lock-in and no notice period. Cancel by replying to any email from us
Run the free scan first. It costs nothing and it tells you whether there is anything here worth watching.
Managed profile
If the profile is missing, we want to publish it for you
A managed UCP profile served on your own domain, generated from your real catalogue and moved to each new spec version as it lands. This one is still being built, so there is nothing to buy and no price to quote yet.
- A conformant profile at /.well-known/ucp on your own domain, no redirect
- Correct Content-Type and public cache headers, held to spec
- Moved to each new UCP version, so negotiation never goes stale
- A one-line proxy rule for nginx, Apache, Cloudflare, WordPress or Next.js
We are not charging for it until the service behind it answers correctly. Leave your domain there and we will come back to you when it is live.
FAQ
Straight answers
What is the Universal Commerce Protocol?
UCP is an open standard for how AI agents discover and transact with a business. A store publishes a JSON profile at /.well-known/ucp describing the protocol version it speaks, the services it exposes and the payment handlers it accepts. An agent fetches that profile, intersects its capabilities with yours and negotiates from there. No profile, no negotiation.
I am on Shopify. Do I need this?
Probably not for the profile itself. We checked: Shopify already serves a conformant UCP profile for its merchants, currently version 2026-08-25. Run the scan anyway, because the other three checks are yours, not Shopify's, and a blocked AI crawler in robots.txt will undo the rest. We will tell you plainly if there is nothing for us to sell you.
Why can't I just publish the file myself?
You can, and if you run your own web server it is an afternoon. Two things make it stick: the spec forbids serving the profile through a redirect, which rules out the simplest workaround on most hosted platforms, and version matching is exact. The spec moved three times in eight months, and a profile pinned to an old version stops negotiating silently.
Will this make AI agents recommend my products?
No, and anyone promising that is guessing. Discovery and conformance are necessary, not sufficient. What we can tell you honestly is whether an agent following the protocol can find you, read your catalogue and price it. Whether it then picks you is down to your products, price and reputation.
What does the scanner actually do to my site?
A handful of GET requests to public URLs: /.well-known/ucp, /robots.txt, /sitemap.xml, /llms.txt, your homepage and up to three product pages. It identifies itself as StoreDeployBot. It does not log in, submit forms or crawl at volume.