How do I test my Shopify store with an AI shopping agent?
Last updated
To test your Shopify store with an AI shopping agent, give a browser-capable assistant your store URL, a shopping need and a budget. Ask it to find a suitable product, choose the right variant and add it to the cart. Keep the pages it used and the exact point where it stopped, then repeat the same task after you make a change.
1. Give the agent a shopping task
Choose a product you sell and a need that its page should answer: a size, an ingredient, a delivery date or compatibility with something the shopper already owns. For example, open ChatGPT Work on web or mobile and paste the task below; approve access to your store if prompted. Its cloud browser can act on websites when available to your account. See OpenAI’s browser guide.
Visit [store URL]. Find a [product] under [budget] for [need]. It must meet [two requirements]. Check the product details, choose the correct size or variant, and add one to the cart. Stop at the cart, before opening checkout. Do not sign in, create an account or place an order. If a fact is missing or a bot check blocks you, stop and say what you could not verify. Report the product, variant, price, pages used and where you stopped.
Use details that a shopper could reasonably ask about and that you can verify yourself. Keep the task unchanged when you repeat it.
2. Try tasks that test different parts of the store
| Shopping task | What to look for | Evidence to keep |
|---|---|---|
| A gift under a budget, needed by a date | Does the agent find price and delivery information, or guess? | The product and shipping pages it used |
| A specific size, colour or pack size | Does the cart contain the requested variant and quantity? | The product selection and cart |
| A product with an ingredient or compatibility requirement | Can the agent read the fact on the page and use it to choose? | The text it relied on and the chosen product |
| A request no product meets | Does the agent explain the mismatch, or silently change the request? | Its answer and the products it considered |
These are task examples, not results from deciq checks.
3. Record the stop before choosing a fix
For each attempt, keep the task, assistant and model, date, pages visited, selected product and variant, and final cart. Save screenshots or the session transcript where available. Record what the agent actually did alongside what it said it did.
Read the page where it stopped. An ingredient list in a picture, an unavailable size and a pop-up over a button call for different changes. Ask which part of the task was unmet and which page should have supplied the answer. If the agent reports a missing fact that you can see, compare its screenshots or quoted page text with the page. If it provides neither, mark the cause as unverified.
After a fix, repeat the same task with the same assistant and model in a fresh conversation. Record any change in stock or price. A successful retry shows what happened on that task; it does not establish that every AI agent can shop on your store.
4. Check discovery separately
A test that starts with your store URL measures what an agent can do once it reaches you. To check discovery, start a fresh AI search with the product need, budget and requirements, without naming your store. Record whether your store appears, the product and link shown, and whether its facts are correct.
Shopify's Agentic Storefronts documentation describes how eligible products reach AI channels and where merchants manage those channels. That is a separate question from whether a browser agent can use your storefront.
What deciq checks
deciq sends eight AI shopping agents to your store, each shopping for a different made-up person. Its report shows where the visits stopped, what the agents could and could not read, and what to change. It tests visits on your store; it does not measure how often ChatGPT recommends you.
Read what a deciq check includes and how to recognise our visits.
Questions
Can I run this test myself?
Yes. Use the prompt above with an assistant that can operate websites, then compare the cart and its answer with your product pages. Start with a task you can check yourself.
Is appearing in ChatGPT the same as passing a shopping test?
No. A recommendation shows that a product appeared in that answer. A shopping test checks whether the agent can use the store to carry out the request. Keep a separate record for each.
Do I need to place an order?
No. The prompt above stops at the cart. If you also want to test payment and order processing, use Shopify's test-order instructions as a separate test. Shopify warns that customers cannot place live orders while a payment provider is in test mode.
Does a successful test mean every agent can buy from my store?
No. The result belongs to that task, assistant, model and date. Other tasks or agents can meet different problems. Shopify's Checkout WebMCP documentation also notes that bot detection can block or deprioritize unidentified browser agents.
Can deciq improve my AI search ranking?
A deciq check reports what happens during shopping visits to your store. It does not measure or promise placement in AI search answers.
Related guides
- Can AI shopping agents buy from my Shopify store?
- How to tell AI visits apart in your store's analytics
- What is Web Bot Auth, and why does Shopify ask bots to sign?
Want a check across eight different shoppers? Join the deciq waitlist.
Join the waitlist