How to Make Shopify Products Discoverable in ChatGPT Shopping

Shopify can make eligible products available to ChatGPT through Shopify Catalog. That is the starting point, not the finish line. Catalog access is a transport decision; query visibility is a selection decision. ChatGPT can receive the structured product data without showing every product for every shopping request, ranking it above competing products, or presenting it exactly as it appears on the product page.

The practical work has five parts:

1. Confirm that the store and products are eligible.
2. Inspect how Shopify Catalog represents the catalog.
3. Make product data specific enough to match real shopping constraints.
4. Strengthen the evidence shoppers use to compare products.
5. Test and measure visibility without treating one chat response as a ranking report.

This distinction matters because a merchant can be technically included and commercially invisible. The catalog may contain the product, but the title may hide the category, the description may omit the material or use case, the correct size may be out of stock, or the product may be less relevant than alternatives for that shopper's request.

Understand the path from Shopify to a ChatGPT shopper

For an eligible Shopify merchant, the main path is:

Shopify product record → Shopify Catalog → ChatGPT product selection → merchant selection → Shopify checkout

Shopify says its ChatGPT Agentic Storefront is a product-discovery channel. A shopper who chooses a product completes the purchase through the merchant's online-store checkout, either in ChatGPT's in-app browser or a new browser tab.

That flow contains several separate decisions:

Do not compress these layers into “ChatGPT SEO.” A visibility problem could be an eligibility failure, a catalog-data failure, a relevance failure, or a competitive failure. Each one needs a different fix.

Pass the eligibility gate before rewriting product pages

Start in Shopify admin → Sales channels → Agentic. Shopify's current ChatGPT requirements say:

• The store must sell to customers in the United States, although the business can be based elsewhere.
• Products must be eligible for Shopify Catalog.
• The merchant must accept Shopify's Agentic Storefronts supplemental terms.
• Terms of service, privacy policy, and return and refund policy must be completed in Shopify settings.

ChatGPT Catalog access is active by default for eligible stores when Allow Shopify to manage for me remains enabled. Merchants who turn that setting off can manage individual channels. If Catalog access has been deactivated, changing a product description will not repair the missing data path.

Product eligibility also matters. Shopify excludes B2B-only products when it can identify them, and products covered by OpenAI's commerce restrictions cannot appear in ChatGPT product-discovery results. A custom B2B setup that hides prices or products through theme logic may not be interpreted the same way as Shopify's native B2B catalogs, so those stores need extra care.

Shopify also instructs merchants to place relevant legal disclosures within the first 6,000 characters of the product description. That is not a writing tip. It is a product-compliance requirement for this channel.

Eligibility checklist

• [ ] The store sells to US customers.
• [ ] The relevant products are published and eligible for Shopify Catalog.
• [ ] ChatGPT has Shopify Catalog access in the Agentic settings.
• [ ] The supplemental terms have been accepted.
• [ ] Terms, privacy, shipping, return and refund policies are complete.
• [ ] The product category is allowed by OpenAI commerce policy.
• [ ] Required disclosures appear early enough in the description.
• [ ] The product URL is public and purchasable without a wholesale login.

Fix failed boxes before working on copy, images or prompts.

Inspect what Shopify Catalog actually sends

Shopify Catalog is the structured product layer used by Shopify's agentic channels. Shopify says it can send titles, descriptions, options, images, price, availability and other key attributes, with price and inventory kept current through the catalog.

For a simple store, the defaults may be correct. For a store with custom product architecture, the storefront and catalog can tell different stories.

Common examples include:

• A short merchandising title appears on the storefront, while the useful category name lives in a metafield.
• A long description contains technical specifications, while the shopper-facing summary lives in a metaobject.
• Related products are grouped through a tag prefix or title delimiter rather than Shopify Combined Listings.
• Internal option names such as CL01 or SZ-A are understandable to staff but meaningless to a shopper or model.
• Several products should be treated as variants of one item, but the grouping method sends them as unrelated listings.

Shopify's Catalog Mapping tool lets a merchant choose sources for the product title, description and category. It also supports custom grouping based on product titles, metafields or tag prefixes, and it lets merchants define the option names shown through Shopify Catalog.

Use mapping only when the default representation is wrong. A complicated mapping is not an upgrade by itself. It creates another configuration that someone must understand and maintain.

Review representative products, not only the bestseller

Preview at least one product from each catalog pattern:

1. A simple product with no variants.
2. A product with size or colour variants.
3. A product with custom metafields or metaobjects.
4. A grouped or combined listing.
5. A sale item.
6. A preorder or backorder item.
7. A product with a category-specific disclosure.

For each product, compare Shopify admin, the live product page and the Catalog preview. The title, description, category, price, availability, images, variants and option labels should describe the same purchasable item.

Catalog work fails when a team optimises the ideal product and leaves the awkward catalog patterns untouched. Those edge cases are usually where grouping, stock and naming errors live.

Improve the listing signals Shopify documents

Shopify now exposes a search preview and Listing quality indicator inside the Agentic section. Its documented listing insights cover five areas:

1. Description completeness
2. Image coverage
3. Verified product reviews
4. Variant and option completeness
5. Store-policy completeness

These are useful because they are merchant-controllable. They are not a complete ranking formula. Shopify separately notes that relevance can also depend on popularity, customer engagement and brand recognition, while an AI channel can rerank Shopify Catalog results using its own logic.

Write descriptions that express purchasing constraints

ChatGPT shopping queries are often combinations of category, use case, attributes and limits:

• waterproof trail shoes for wide feet under $140
• linen curtains that block some light but are not blackout
• fragrance-free moisturiser suitable for dry skin
• carry-on backpack that fits a 16-inch laptop

A description such as “premium quality for modern lifestyles” contributes almost nothing to those decisions. It names no material, fit, compatibility, measurement, use condition or limitation.

A useful product description should make the following facts explicit when they apply:

• What the product is
• Who or what it is designed for
• Materials or ingredients
• Dimensions, weight or capacity
• Compatibility
• Fit or sizing behaviour
• Use conditions
• Care requirements
• Included and excluded components
• Safety or legal disclosures
• Meaningful limitations

This is not permission to stuff attributes into unreadable prose. Keep a clear opening summary, then use structured specifications for details. The goal is to help a shopper establish fit, not to repeat the product category twenty times.

For large catalogs, the EcomCX product-content workflow can help teams define fields and review rules before generating or revising descriptions in bulk.

Build image coverage around decisions

Shopify measures image coverage because different views help an AI channel represent a product in different contexts. More images are useful only when they remove uncertainty.

For a physical product, the set might include:

• Clean primary image
• Alternate angle
• Scale or in-use image
• Material or texture detail
• Variant-specific image
• Dimensions or fit reference
• Packaging and included components

Do not upload six near-identical angles and call the catalog complete. A shopper choosing a bag cares whether a laptop fits. A shopper choosing a table cares about scale. A shopper choosing clothing cares how the stated size relates to the pictured fit.

Preserve trustworthy review evidence

Shopify's Listing insights use reviews verified by trusted sources. OpenAI says ChatGPT may display model-generated summaries based on reviews from public websites, but also warns that those reviews and ratings are not verified by OpenAI.

That creates two responsibilities for the merchant:

• Keep first-party and approved review integrations accurate and policy-compliant.
• Make product claims defensible without relying on a generated review summary.

Do not manufacture review text, ratings or counts to fill a signal. False social proof creates a trust and policy problem, not a discoverability strategy.

Complete policies as product evidence

Shipping, returns and refund policies help an AI channel understand whether the store presents a credible buying path. They also answer constraints that can affect the shopper's decision.

The policy should state practical details: destinations, processing times, delivery estimates, return window, condition requirements, return costs, exclusions and refund timing. A policy titled “Easy returns” with no rules is not useful evidence.

Treat every variant as a purchasable record

Variant errors are especially damaging because the product can look correct at parent level while the purchasable record is wrong.

OpenAI's product-feed specification is not a setup requirement for Shopify merchants using Shopify Catalog. It is still a valuable data-quality reference. The specification treats each product or variant as a row with a stable ID and documents fields for title, description, URL, image, price, availability, identifiers, grouping and variant attributes.

The useful principle is simple: if a shopper can buy it as a distinct SKU, its data must remain internally consistent.

Prioritise products where variants drive the buying decision: apparel, footwear, furniture finishes, device capacity, bundles, beauty shades and compatibility-based accessories.

Understand what agents.md and llms.txt can and cannot do

Shopify automatically serves three agent-discovery URLs:

• /agents.md
• /llms.txt
• /llms-full.txt

Shopify calls /agents.md the canonical agent-discovery URL. By default, the files can expose store context such as the store name, URL, sitemap, policies and discovery endpoints. A theme can customise them through Liquid templates.

The critical limitation is in Shopify's own documentation: agent-discovery files are separate from Shopify Catalog and do not replace it.

The same separation applies to web crawling. Shopify Catalog is the primary path for activated agentic channels, while AI crawlers may also discover public pages through the open web. robots.txt can provide crawler directions, but blocking a crawler does not turn off Catalog syndication. Turning off Catalog access does not guarantee that a public product disappears from open-web results.

Treat these as different surfaces:

Customise an agent file when the default store context is incomplete or wrong. Do not start there when price, availability, category or variant data is broken.

Test discoverability with a query matrix

Shopify's Catalog search preview should be the first diagnostic tool. It shows raw Shopify Catalog search results, identifies products that may rank for a query, and provides Listing insights. Shopify explicitly says the preview is directional because agentic channels can rerank results.

Use it to answer two questions:

1. Does Shopify Catalog understand which of my products are relevant?
2. Which controllable listing gaps does Shopify identify?

Then run a small set of ChatGPT checks to observe the customer-facing experience. Do not test only the exact product name. Branded queries confirm recognition; they do not prove discovery.

Build queries from real buying constraints

For each priority product family, create six query types:

Record the date, country, account state, query, result format, products shown, merchant shown, price, stock status and any inaccurate details. Account context matters because OpenAI says ChatGPT can consider the shopper's query, memory and custom instructions.

One observation is not a trend. Repeat the same controlled query set monthly and after material catalog changes. Treat the result as an observation of a variable system, not a permanent position.

Fix discoverability problems in the right order

Use this sequence so cosmetic work does not hide a system failure.

1. Access and eligibility

Confirm Shopify Catalog eligibility, ChatGPT access, US market availability, accepted terms, policies and allowed product category.

2. Transactional integrity

Correct broken URLs, stale prices, wrong currency, inaccurate availability, missing preorder dates and checkout failures. ChatGPT may receive updates with some delay, so record when the source changed before judging the result.

3. Product identity and structure

Fix category, brand, stable SKU or identifier, product grouping, variant labels and duplicate listings. If Shopify Catalog cannot identify the purchasable item cleanly, richer copy will not repair the structure.

4. Decision information

Add factual attributes, dimensions, materials, fit, compatibility, inclusions, exclusions and limitations. Use Catalog Mapping when the correct source already exists in a metafield or metaobject.

5. Evidence and confidence

Improve useful image coverage, verified reviews and store policies. Make sure those signals describe the same product and buying terms shown at checkout.

6. Query testing and measurement

Run the same query matrix in Shopify preview and ChatGPT. Log visibility and data errors, then connect those observations to sessions and orders.

Measure business impact, not screenshots of a carousel

Shopify documents Agentic Storefront reporting for:

• Sales
• Orders
• Online-store sessions
• Online-store conversion

These reports are available by AI channel in the Agentic section, although Shopify says agentic performance analytics are not yet available for headless stores. Sales combine referral-based purchases and supported direct-checkout activity.

Track four stages separately:

Do not claim that a description rewrite caused an order because the product appeared once after the change. Price, stock, reviews, competitors, shopper context and channel behavior may also have changed.

A useful monthly review asks:

1. Which priority queries gained or lost relevant products?
2. Which listings still have quality warnings?
3. Which surfaced details are wrong or stale?
4. Which ChatGPT sessions reached product pages?
5. Which product families produced orders or assisted research?
6. Which catalog change should be tested next?

The EcomCX ecommerce AI measurement guide provides a broader framework for connecting channel activity to business outcomes without inventing attribution certainty.

Know what you cannot guarantee

No merchant can guarantee placement for a ChatGPT shopping query.

OpenAI says product results depend on the query and context, and that not every available product will be shown. It can consider structured metadata, price, reviews, ease of use and policy restrictions. Merchant selection can consider availability, price, quality and whether a seller is the maker or primary seller.

There are also presentation limits:

• ChatGPT may simplify product titles and descriptions.
• Review summaries can be generated from public sources.
• Price or shipping updates may take time to appear.
• A product selected in Shopify Catalog preview may be reranked by ChatGPT.
• Personal context can change the result.
• Popularity, engagement and brand recognition are not fixed through one catalog edit.

This uncertainty is a reason to build a repeatable operating process, not a reason to ignore the channel.

A practical 30-day plan

Week 1: Verify the route

Confirm eligibility, Catalog access, policies and product restrictions. Select 20 commercially important products across representative catalog structures.

Week 2: Repair the records

Review Catalog Mapping, categories, titles, descriptions, grouping, variants, prices, stock and images. Correct system errors before expanding copy.

Week 3: Test buyer language

Create six constraint-led queries for each priority product family. Run them in Shopify Catalog preview, record listing insights and spot-check the same patterns in ChatGPT.

Week 4: Measure and choose the next fix

Review ChatGPT sessions, orders, sales and conversion where available. Compare the results with the query log. Choose one catalog hypothesis for the next cycle instead of rewriting the whole store.

The final standard is simple: the product record should let a machine answer the same questions a careful shopper asks before buying. Eligibility opens the door. Accurate, specific and competitive product evidence gives the product a reason to be considered.