AI visibility audit: how to check what AI tells shoppers about your store

When a shopper asks ChatGPT what to buy, it names a few products and says why. An AI visibility audit tells you whether your products are among them, what the AI says about them, and where it got that information.

You can do it yourself in an afternoon, with a spreadsheet and free AI accounts. Here's how, step by step.

Before you start

You'll need three things:

• A spreadsheet. Give it these columns: question, AI assistant, were we named?, what it said, was it right?, and websites it linked to.
• Free accounts for ChatGPT, Gemini, Claude and Perplexity. Google's AI Mode, inside Google Search, works without one.
• Your three best-selling products. You'll check these closely in Steps 4 and 5.

Step 1: Write down what shoppers ask

Start with the questions real customers already ask you. Look in your support emails and chats, your reviews, and the reasons people give when they send something back.

Turn them into about 10 questions, written the way a shopper would ask an AI. Use four kinds:

• Broad questions: "What's the best [product] for [situation]?" This is where the AI picks between stores.
• Questions about one product: "Does [your product] run small?" or "What is [your product] made of?"
• Questions about your store: "How long does shipping take from [your store]?" and "What's [your store]'s return policy?"
• You against a competitor: "[Your brand] or [competitor], which is better for [situation]?"

Write them as full sentences, with details. Shoppers don't type short keywords into AI. Google says questions in its AI Mode are "two or three times the length of traditional search queries."

Step 2: Ask each AI assistant

Ask every question in ChatGPT, Gemini, Claude, Perplexity and Google's AI Mode.

A few rules keep the results fair:

• Start a new chat for every question. Earlier questions in the same chat change the answer.
• Use a temporary chat, or turn memory off. Otherwise the AI may remember that you asked about your own brand before.
• Ask your broad questions twice, on different days. AI answers change from one time to the next.

For each answer, fill in one row of your spreadsheet:

• Were you named?
• What did it say about you?
• Was it right? Check the price, the sizes, the shipping and the returns.
• Which websites did it link to?

A wrong answer can be worse than no answer. If an AI tells a shopper your return window is 14 days when it's really 30, that shopper may never find out the truth.

Step 3: See where the answers come from

Copy every link the AI assistants gave you into a second tab. Sort each one into a group:

• your own website
• review websites
• "best of" list articles on other websites
• Reddit and other forums
• videos
• other stores, including your competitors

This shows you where your category gets decided. For broad "best" questions, AI leans heavily on list articles written by other people. One study found they made up 43.8% of the pages ChatGPT linked to, more than any other kind of page.

Step 4: Read your product page the way AI does

Many AI assistants, including ChatGPT, Claude and Perplexity, read your product page as plain text. They often skip anything that only shows up after the product page finishes loading.

Hidden code on the product page, called schema markup, won't fix this on its own. A large study by Ahrefs found that adding it didn't make AI name pages as a source any more often. What counts is the text a shopper can see.

Open the product page for one of your best sellers. Check that each of these is written on the product page, as words:

• what the product is for, and who it's for
• what it's made of
• size and fit
• the price, and whether it's in stock
• how long shipping takes, and what it costs
• how returns work

Then look for three common gaps:

• Facts that are only in pictures. A size chart saved as an image can't be read as text.
• Reviews that load late. If your reviews pop in a moment after the product page opens, many AI tools never see them.
• Shipping and returns only in a footer link. A link that just says "Refund policy" doesn't tell AI what the policy is.

Step 5: Check what Shopify shares with AI

AI shopping assistants, like the ones in ChatGPT, Copilot and Gemini, don't read your product page. They get product details from what you've saved in your Shopify admin. So your product page can look complete while AI gets something quite different. Here's a real example.

In your Shopify admin, open the same best seller and check:

• The Description box is filled in. Some store designs show a description that's saved somewhere else. If this box is empty, AI shopping assistants get a blank description.
• Sizes and colors are set up as options, not only mentioned in the description.
• Your shipping and return policies are written out in full, under Settings, then Policies. A placeholder like "." or "[INSERT RETURN ADDRESS]" is as good as missing.
• If you use Shopify's Knowledge Base app, each answer sits under a specific question a shopper would ask. In one store we checked, AI found the answers only when it asked a narrow question. Here's that example.

Step 6: If your website is custom-built, check the front door

Skip this step if your store uses a regular Shopify theme.

If your website was built separately and connects to Shopify behind the scenes, type these two addresses into your browser, with your own website in place of yourstore.com:

• yourstore.com/agents.md
• yourstore.com/.well-known/ucp

If either one says "page not found," AI shopping assistants that start from your website can't find the files that tell them how to shop your store. Here's what that looked like for one brand.

Step 7: Add it up

From your spreadsheet, work out four things:

• How often you were named. Out of all your questions, how many answers mentioned your brand? Work this out for each AI assistant.
• How often the answer was right. Count the answers that got your price, sizes, shipping or returns wrong.
• How much comes from your own website. What share of all the links went to your website?
• The gaps you found in Steps 4, 5 and 6.

Keep the spreadsheet. Ask the same questions again once a month, so you can see what changes.

What you can't control

AI answers change from day to day, and the same question can get two different answers. Each AI assistant picks its sources in its own way, and none of them publish exactly how. An audit tells you where you stand today. It can't promise you a spot in tomorrow's answer.

Want a head start?

Run the free check. It does much of Steps 4, 5 and 6 for you, on one of your product pages.