Can you measure whether AI assistants recommend your Shopify store?
A shopper asks an AI assistant where to buy something you sell. Does it name your store?
You can measure part of that.
You can ask ChatGPT, Perplexity, Claude, and Gemini the same shopping question. You can record whether they named your store, which brands appeared instead, and which sources they cited.
But that result should not become one neat “AI visibility score.”
Say your store is named in 12 of 36 answers. That is useful. It does not tell you why the other 24 left you out. The gap might be in product data, product pages, policies, crawler access, or the question itself.
Descriva keeps those jobs separate. It has one score: AI readiness, out of 100. It reports AI mentions as counts, question by question.
That distinction helps when you need to decide what to fix.
A readiness score and a mention count answer different questions
AI readiness asks whether your store gives AI assistants and shopping agents usable information.
Descriva reads the store and checks five areas:
- Product data
- AI shopping agents
- Live pages
- Trust and policies
- Crawler access
The score comes straight from the store. It does not need an AI call.
A product with no image, a price of 0, or a product hidden from search engines can create a problem before an assistant considers recommending it. So can a password-protected store or crawler rules that block access.
AI Mentions checks something else. Descriva asks the four assistants questions a shopper might ask, such as:
“Where can I buy lightweight waterproof hiking jackets?”
It then records:
- Which assistant named your store
- Which assistant left it out
- Which competing brands appeared
- Which sources the answer cited
You might see that Claude named your store for a product question while Gemini did not. Or that an assistant describes a product correctly but does not connect it to your brand.
Neither finding is a score. They are answers you can inspect.
Shopify supports product discovery through several AI channels. It also attributes orders that come through those channels in Shopify admin. That is useful for measuring downstream orders. It is different from measuring whether a general AI answer mentioned your store at all. Shopify’s agentic storefront documentation covers both the available channels and order attribution.
A single number hides the work you need to do
Say your store sells refillable cleaning products.
An assistant may leave you out because your collection page never explains how the refills work. It may be because product descriptions do not say which surfaces they are safe for. It may be because shipping regions are unclear.
It may also be that a competitor has a clearer page for the exact question.
Those are different problems. They need different fixes.
A mention count tells you where to look. The readiness score helps identify store issues you can fix directly.
Descriva’s dashboard shows the readiness score, one next step, and AI crawler visits and mention checks from the last 7, 30, or 90 days. The useful part is not watching a number move. It is finding the next piece of work worth doing.
Ask questions your customers would actually ask
The question shapes the answer.
“Best skincare brands” is broad. It may not resemble the way your customers shop. A more useful question could be:
“Where can I buy fragrance-free moisturiser for dry skin in Canada?”
Or:
“What are good gifts for someone who makes espresso at home?”
Descriva starts with product questions. Starter asks one question per product. Pro asks three, including whether assistants know the brand.
You can add questions of your own too:
- Starter includes three
- Pro includes 10
- Done for you includes 25
Use questions from support tickets, product reviews, DMs, and sales conversations. Keep them specific enough that a real shopper could have asked them.
You can also add up to 10 competing brands. Descriva shows how often each assistant names your store and those competitors in the same answers.
If an assistant names a competitor, read the answer and its cited sources. You may find that the competitor has a category page, product guide, or policy page that answers the shopper’s question more directly.
That does not prove why the assistant chose them. It gives you a concrete page to compare.
Check what Shopify Catalog can receive
A polished product page is not the whole picture.
Shopping agents can depend on the product information Shopify Catalog receives. Descriva’s Agent view shows what each product provides through Shopify Catalog and flags products that miss basic requirements:
- The product is draft or unlisted
- The product has no image
- The price is 0
- The product is not published to the Online Store
- The product is hidden from search engines
- The store is behind its password page
You can check several of these yourself in under a minute. Open a product in Shopify and review Status, Sales channels and apps, Search engine listing, Media, and Pricing.
If a product is not published to the Online Store, rewriting its description will not solve that problem.
Shopify’s own guidance also points merchants to product availability, store details, policies, and shipping information when preparing for AI channels. Its product eligibility guidance is here.
Read the answers your store gives agents
A shopping agent needs basic buying information.
Can the store ship to the shopper? What will shipping cost? When will it arrive? What is the return process? How can the shopper get in touch?
Descriva shows, word for word, what your own agent tool tells shoppers’ AI agents about:
- Returns
- Shipping cost
- Delivery time
- Shipping regions
- Contact details
When an answer is missing, Descriva links to the place to add it as an FAQ in Shopify’s free Knowledge Base app. When you return, Descriva asks again.
This is worth checking even if you have policy pages. A return policy may state the return window but omit who pays return shipping. A shipping page may list countries but not delivery times.
Shopify says its Knowledge Base app can help AI platforms represent a business accurately. It also says the app does not affect how often a store appears in AI-platform results. Accuracy and appearance frequency are separate jobs. Shopify’s Knowledge Base documentation says so directly.
Fix the gap, then review every change
Once you know what is missing, you need a practical place to fix it.
In Descriva’s Optimize tab, each product opens in one popup. You can work on its description, search listing, title, image alt text, FAQ pack, and specs pack.
You can draft copy with AI or write it yourself. Nothing publishes until you press Publish. History keeps what each change replaced.
That review matters. A product draft may need a corrected measurement, material, compatibility detail, or shipping claim before it belongs on a live page.
Descriva can also add Product, FAQ, and Organization schema, plus breadcrumbs, through a theme app embed.
We deliberately ship Descriva without permission to write to a theme. Fewer permissions means less for a merchant to trust. The trade-off is that a custom `llms.txt` file comes as a template for you to paste. Descriva then reads the live URL instead of assuming the paste worked.
Shopify already serves a default `/llms.txt` for stores. For most stores, Shopify says its managed file is sufficient. A custom file is not automatically required. Shopify’s developer documentation explains the default setup.
Use the result as evidence, not a promise
AI answers can change. Models change, sources change, and the same question can be framed differently.
A mention is not a promise of traffic or sales.
It is still useful evidence. If ChatGPT names your store in 3 of 9 answers after never naming it before, inspect the questions and sources. If your store is repeatedly absent from questions where its products should fit, inspect the product, collection, policy, and agent information behind those answers.
Use AI readiness for parts of the store you can check and improve directly. Use mention counts to see how assistants answer real shopping questions.
If you want to start without installing anything, run the lunalink.ai scan. It reads public store pages with no login and no app access.