Can ChatGPT and Perplexity tell which Shopify product variant a customer should buy?
A customer might ask ChatGPT: “Which version of this jacket should I buy for wet winter walks?”
That question is not really about a product title. It is about size, material, warmth, waterproofing, colour, price, and whether the right option is in stock.
For a Shopify store, the useful question is simpler: can an AI assistant see enough of your variant data to make a sensible match?
Sometimes, yes. But only when the store gives it something clear to work with.
AI assistants can compare constraints
ChatGPT can help shoppers research products using constraints such as brand, size range, budget, performance, comfort, style, and price. Its shopping research can compare product attributes side by side.
Perplexity can also recommend products from natural-language shopping questions.
Neither assistant knows what a shopper should buy in the way a skilled shop assistant might. They cannot feel a fabric, ask follow-up questions forever, or know a customer’s fit preferences unless the customer gives them that information.
What they can do is match a request against available product data.
Say your store sells a running shoe in three versions:
- Road, with more cushioning
- Trail, with deeper grip
- Waterproof trail, for wet conditions
A shopper asks for a shoe for muddy winter runs. If the assistant can read those differences, it has a basis for recommending the waterproof trail version. If every option is called only “Black / 8” or “Blue / 9,” it has much less to go on.
The product page may look clear to a person who has clicked around. That does not mean the underlying variant information is clear enough for an AI system.
Shopify sends variant information to AI channels
Shopify Catalog sends product information to AI channels, including titles, descriptions, options, images, prices, availability, and other structured attributes.
Shopify says this information is updated continuously across AI channels. That gives systems such as ChatGPT information that can distinguish options like size, colour, and style.
Shopify also says complete variant data can help AI channels match purchase intent. Its listing-quality check looks at variant count, option names, and whether variants are in stock.
That is a useful standard for merchants: a variant should not merely exist in Shopify. It should say what it is.
Compare these two option setups.
Hard for a system to interpret:
- Option name: Variant
- Values: Standard, Plus, Pro
More useful:
- Option name: Insulation level
- Values: Lightweight 80g, Warm 150g, Cold-weather 250g
The second version does not guarantee a recommendation. But it gives an assistant facts it can compare to a shopper’s request.
“Should buy” depends on more than option names
A clean variant name is a start. It is not the full job.
If a customer asks which size they should buy, the assistant needs a size guide or sizing facts. If they ask which coffee grinder suits espresso, it needs to know what differs between burr sets or settings. If they ask which bundle is best for a gift, it needs the contents of each bundle.
Put the deciding facts where they are easy to find.
That usually means:
- Clear product titles
- Option names that describe the real choice
- Variant values with plain labels
- Product descriptions that explain the trade-offs
- Specification lists for measurable details
- FAQs for recurring buying questions
- Accurate price and availability
A product description can explain the overall product. The variant information needs to explain the choice within it.
For example, a blanket product might say it is made from wool. That helps with broad discovery. But a shopper choosing between variants may need to know whether one is a throw, a single-bed blanket, or a larger size.
Do not make them infer that from an internal SKU.
Check one product in under a minute
Open one product in Shopify admin and look at the variants section.
Ask these questions:
- Would a first-time shopper understand each option name?
- Do the values describe a real difference, or just use internal labels?
- Is each variant’s availability correct?
- Does the page explain which version suits which use?
- Could someone tell the difference without looking at product photos?
Start with a product that has several variants and gets regular traffic. That is where unclear data creates the most buying friction.
If you find “Small,” “Medium,” and “Large,” that may be fine. But make sure the page also gives dimensions or a size guide. If you find “Model A,” “Model B,” and “Model C,” add the difference in plain language.
A catalogue can look complete when it is not
When we built our own llms.txt builder, we found a quiet catalogue problem in Shopify’s `/products.json` endpoint.
It returns at most 250 products per page.
If you read only the first response, a larger catalogue can look complete when it is not. The missing products do not announce themselves. You need to handle pagination to read the rest.
Variant work has a similar risk. You can inspect a few polished products and assume the catalogue is in good shape. Then an older collection, a seasonal item, or a product imported from another system has vague options and missing details.
Audit beyond your best sellers.
Look for products where variant names are generic, descriptions are thin, or availability has not been checked recently. Those are not only AI search issues. They are ordinary product-page issues too.
ChatGPT is a discovery path, not your Shopify checkout
For Shopify merchants, ChatGPT is currently positioned mainly as a discovery and referral channel. Shopify says customers who find products through ChatGPT complete checkout on the merchant’s online store, either in ChatGPT’s in-app browser or a new tab.
OpenAI also says Shopify product data is already integrated into ChatGPT through Shopify Catalog, with no extra work required from individual merchants for that integration.
That does not mean there is no work left for the merchant.
The connection may exist, but the quality of what an assistant can compare still depends on the information in your catalogue and on your storefront.
A shopper may arrive on your product page after an AI recommendation and still need to choose a size, colour, or configuration. The page should finish the job without making them start the research again.
What Descriva checks
Descriva audits how AI search systems can read your Shopify store. That includes the product facts, copy, structured data, and gaps that make it harder to understand what you sell.
It can generate entity-first product copy, FAQs, specification lists, and JSON-LD schema. “Entity-first” here means starting with the actual thing: what the product is, who it is for, what it is made of, and how its versions differ.
That is more useful than adding vague phrases about being “perfect for every occasion.”
If a mug comes in 10oz and 16oz versions, say what each holds. If a skincare product comes in fragrance-free and scented variants, name the scent and explain the difference. If a chair has fabric grades, list the materials and care needs.
The goal is not to write for a bot. It is to make the product decision legible to both a person and the systems helping them search.
Start with the choices customers ask about
Look through your support messages, returns, and product reviews. Find the questions that come up before purchase.
They may be simple:
- Which size fits a child?
- Is this version compatible with my device?
- What is included in each bundle?
- Which colour is closest to the photo?
- Is the premium version more durable, or just larger?
Those questions show where your product data needs more detail.
Give each answer a clear home on the product page or in a useful FAQ. Then make sure the variant names and availability match what the page says.
That is a practical first step whether a customer finds you through ChatGPT, Perplexity, Google, or a direct visit.