How Do You Test Whether AI Agents Can Actually Use Your Shopify Store?

November 17, 2026

By Steve Merrill · November 17, 2026

How Do You Test Whether AI Agents Can Use Your Shopify Store?

Run five checks: confirm AI bots can crawl your pages, fetch a product page as raw HTML, validate your Product structured data, prompt AI assistants about your products, and walk a guided purchase with a browsing agent. Each check takes minutes, and together they show exactly where an agent gets stuck.

Most store owners skip this. They assume a good-looking store works for everyone, including software.

It doesn't. Agents read code, not design. Here's the test I'd run on any store.

Step 1: Can AI Bots Even Reach Your Pages?

Open yourstore.com/robots.txt in a browser. Look for rules that block OAI-SearchBot, ChatGPT-User, PerplexityBot, or Googlebot from product and collection pages.

OpenAI documents its crawlers on its bots page, and Perplexity does the same in its crawler documentation. The two companies publish separate user agents for search and for other uses, so read both before you decide what to allow.

Shopify lets you edit this file through a robots.txt.liquid template, covered in the Shopify developer docs. If an app or an old edit blocked these bots, you'll see it here.

A blocked bot can't read your catalog. Nothing else on this list matters until this passes.

Step 2: Does a Product Page Make Sense Without JavaScript?

Fetch a product URL with curl and read what comes back. Your product title, price, availability, and variant names should all be in that raw HTML.

curl -s https://yourstore.com/products/your-product | grep -i "price"

If the price only shows up after a script runs, some crawlers will never see it. Many AI crawlers fetch pages without running a full browser. Blank.

I ran this check on a store with a beautiful custom theme. The variant selector was built entirely in JavaScript, so the raw page listed no sizes and no stock status. A human saw a clean product page. A crawler saw almost nothing.

Step 3: Is Your Product Structured Data Correct?

Paste product URLs into Google's Rich Results Test and the Schema Markup Validator. You want a valid Product with name, image, price, currency, and availability.

Google's Search Central guidance on Product markup lists the properties it recognizes. The schema.org Product definition is the shared vocabulary behind it.

Then compare the markup to the visible page. Does the structured price match the price on screen? Does "InStock" match what a shopper sees?

Mismatches are common after a sale or a variant change. An agent that finds two different prices has to pick one, or skip you.

Step 4: Do AI Assistants Mention Your Store?

Open ChatGPT, Perplexity, and Gemini. Ask the questions a buyer would ask, using your category and your price range.

Write down five prompts before you start. Keep them the same each time you retest.

  • "Best [your product type] under [your price]"
  • "Where can I buy [your exact product name]?"
  • "What's the return policy for [your brand]?"
  • "Compare [your brand] with [a competitor]"
  • "Is [your product] in stock in [a size or color]?"

For each answer, log three things: whether you were named, whether your site was cited, and whether the details were correct. Wrong details are the most useful result. They point straight at the page the model read badly.

Step 5: Can an Agent Complete a Purchase?

This is the test that matters most. Give a browsing agent a task: pick a specific size and color, find the shipping cost to a given zip code, locate the returns policy, and get to checkout.

Watch where it stalls. Common stall points:

  • A popup that covers the add-to-cart button
  • A variant selector that doesn't update the price
  • Shipping costs that only appear at the final checkout step
  • A returns policy buried in a PDF
  • A CAPTCHA or forced account creation before payment

Every stall is a lost sale from a buyer who never saw your store. The agent moves to the next one.

Run the task three times. Agents behave differently run to run, so one pass tells you little.

How Do You Score and Fix What You Found?

Score each step pass, partial, or fail, and fix in order. Bot access first, raw HTML second, structured data third. The prompt tests and purchase run tell you whether those fixes worked.

Keep a simple log with the date, the step, the result, and the page URL. Rerun after every theme change or app install. Apps are the usual culprit when a passing store starts failing.

Here's the bottom line: if an agent can't read your price, find your policies, and reach checkout, it can't buy from you.

Doing this by hand across a full catalog takes time. If you'd rather have a team run it and fix what breaks, see how WRKNG Digital approaches agentic commerce for Shopify stores.

Frequently Asked Questions

Do I need special tools to test AI agent readiness?

No. A browser, curl, Google's Rich Results Test, the Schema Markup Validator, and free accounts on ChatGPT, Perplexity, and Gemini cover the full test. The most useful tool is a written log of what each check returned.

Should I block AI crawlers on my Shopify store?

Only if you want to stay out of AI answers. OpenAI documents separate crawlers for search and training, so you can allow OAI-SearchBot and ChatGPT-User while making a different choice about GPTBot. Make that decision on purpose.

How often should I rerun the test?

Rerun it after every theme change, app install, or checkout change, and at least once a quarter. Apps and theme updates are the most common reason a store that passed last month fails today.

What is the most common failure?

Missing or mismatched product data. Price, availability, or variant details in structured data that don't match the visible page give an agent a reason to skip the store or quote wrong information.

Back to Blog