Agent readiness

Being listed in an answer is half of AI shopping. The other half is whether an assistant has what it needs to recommend a specific product, and whether a shopping agent can then open your site, find the product, read its price and reach the cart. The Readiness tab checks both and turns every gap into a fix that leads somewhere you can act.

#Where to find it

Open Analytics › E-commerce, then the Readiness tab. If you have no products yet, the tab asks you to add or import some first; the agent test still works, because it needs only your website.

From top to bottom the tab shows:

1. Three tiles: Readiness score, Standing and Surfaces open to you.
2. The AI shopping feed and the Could an AI agent shop on your site? test, side by side.
3. One card per check family, the Which AI shopping surfaces can you enter? table, and What to fix, in order.

#The readiness score

The score runs from 0 to 100 and is made of five families of checks:

Crawler access weighs heavily on purpose: if the crawler that writes the answer is blocked, nothing else on the page helps. It is also the cheapest fix.

#Status of each check

Share-based checks (the feed fields, reviews) use the same thresholds:

#Checks that could not be run

A check Rixe could not run — a page that did not load, a robots.txt that could not be read, a column your file does not have — is marked Not measured and left out of the score, never counted as a failure. The score is calculated over the points that could be measured, and the Readiness score tile tells you how many of the 100 points that was. When something was left out, a notice at the top says how many points' worth of checks could not be run.

#Standing

#Market reality

Merchant programmes for AI shopping open country by country. The Which AI shopping surfaces can you enter? table lists each surface with Open or Closed, the reason, and the date the fact was verified from the programme's own documentation. A closed surface is not a shortcoming of yours.

The default table, unless your market defines its own:

"Needs your own product page" means the surface cannot read marketplace pages: if you only sell through a marketplace, you are effectively absent there.

#What to fix, in order

This list is the deliverable. Every failing or incomplete check appears with the points it costs you, the most expensive first. The Where link takes you to the place the gap is closed:

The footnote names the product page that was measured. If none could be found, add a domain to your brand.

#The agent test

Could an AI agent shop on your site? walks your site the way ChatGPT's, Perplexity's and Claude's shopping agents would. It announces itself with those agents' own user-agents (ChatGPT-User, OAI-SearchBot, Perplexity-User and Claude-User). Nothing is clicked and no cart is filled.

#Running it

1. Make sure your brand has a website set.
2. Press Run the agent test. Owners, admins and members can run it.
3. The result appears on the card within about a minute.

One test per site can run every ten minutes; pressing the button sooner tells you the last test ran a few minutes ago.

#The six steps

Each step reads Passed, Blocked or Not reached. The Agent readiness score counts only the steps the test could reach, so a step it never got to says nothing about your site. If no step could be measured at all, the card says the site may not have answered. The footnote shows the date and the address tested.

#Common questions

My score went up although I fixed nothing. A check that could not run before may have run this time, changing which points the score is calculated over. Look at how many of the 100 points could be measured.

ChatGPT shows as Closed for me. Its merchant programme is limited to the United States. Google AI Mode accepts every country.

Will the agent test place an order? No. It only reads pages: nothing is clicked and nothing is added to a cart.

#Related pages

• E-commerce overview
• AI shopping feed
• Site health
• AI traffic and crawlers