If your Shopify products are not appearing in ChatGPT Shopping, do not start by rewriting every product description or adding a long list of AI crawler rules. First identify which product-discovery path you are troubleshooting.
For eligible Shopify stores, product data can be shared with ChatGPT through Shopify Catalog. Shopify says no extra merchant action is required for eligible products to be discovered through that catalog path, although eligibility, channel settings, product eligibility, and data mapping still matter. ChatGPT also has web-search and user-requested browsing systems, which are different from a Shopify Catalog product feed. See Shopify's ChatGPT selling requirements and Shopify Catalog product discovery.
That distinction changes the troubleshooting process. A crawler being able to fetch a product page is not the same as a product being eligible in a shopping catalog, and neither guarantees that a product will be selected for a particular shopper's question.
This guide walks through the checks in order, from eligibility to product information and real recommendation tests.
1. Define what "not showing" means
Before changing anything, describe the symptom precisely. These cases have different causes:
- The store or products are unavailable to the channel. Start with Shopify Catalog eligibility, sales-channel settings, and product status.
- A product page can be opened, but the product is absent from a shopping answer. Check the product's data and eligibility, then test whether the prompt actually matches the product.
- A product appears for some prompts but not others. This may reflect differences in shopper intent, price range, requirements, availability, geography, and the answer generated at that time. It is not by itself proof of a technical problem.
- ChatGPT can cite the store website in web search, but does not show a product card. A web citation and a shopping product result are separate outcomes.
- A product appears with incorrect information. Investigate source data, catalog mapping, variant relationships, price, availability, and the timing of updates.
Record the exact prompt, market, date, and result before making changes. A screenshot of one answer is useful evidence, but not enough to diagnose a store-wide issue.
2. Check Shopify Catalog and channel eligibility first
For a Shopify merchant, start in Shopify Admin rather than with a generic AI-readiness checker.
- Open Sales channels > Agentic and review the available channel status.
- Open the ChatGPT channel and confirm that Shopify Catalog access has not been disabled for the store or relevant products.
- Review any eligibility warnings shown in the admin.
- Confirm that the store meets the channel's current requirements, including selling to customers in the United States for the ChatGPT Shopify channel described in Shopify's current documentation.
- Confirm that required store policies are complete. Shopify currently lists Terms of Service, Privacy Policy, and Return and Refund Policy among the requirements for selling through ChatGPT.
- Check that the products are eligible for Shopify Catalog and are not excluded by product or commerce policies.
The exact requirements can change and may vary by channel, country, product, and selling setup. Use the live Shopify guide to selling on ChatGPT as the source of truth rather than relying on an old checklist.
One important nuance: eligibility is necessary for a given catalog path, but it is not a promise that every eligible item will appear for every query. Shopify explicitly says that channel eligibility does not guarantee a particular placement in a specific AI answer. Each channel decides what to show for the shopper's request. See Shopify's agentic storefront overview.
3. Verify the data Shopify is actually sending
An online product page may look complete to a person while the structured catalog data contains missing, stale, or incorrectly mapped fields.
Review a representative sample of products, not just the homepage. For each product, compare the Shopify product record with the information you expect an AI shopping channel to receive:
- Product title and description
- Product URL and primary image
- Price and currency
- Availability and inventory state
- Product type, category, options, and variants
- SKU, GTIN, MPN, or other identifiers where relevant and available
- Brand and manufacturer information where applicable
- Shipping, returns, and other store policies where the channel uses them
In Shopify, inspect the Catalog eligibility details and the product data mapping page where available. Shopify notes that custom fields, metafields, metaobjects, title conventions, and custom grouping logic can be reasons to review data mapping. If the authoritative title or description is stored in a custom field, do not assume that the channel is using the field you intended. See Shopify Catalog product data mapping.
Also check the actual product record after any change. A correct storefront display does not prove that every external feed or catalog source has already refreshed. Record when the change was made and when you rechecked it.
4. Check whether product descriptions are interchangeable
Many ecommerce catalogs contain descriptions that differ in brand name but repeat the same generic claims: "premium quality," "designed for everyday use," or "perfect for any occasion." Those phrases may sound polished, but they do little to explain which product fits a specific requirement.
It is important to be precise here: there is no publicly documented universal ChatGPT Shopping rule that says duplicated descriptions are automatically demoted. Do not claim that unique copy guarantees recommendations. The commercial issue is more direct: if a description contains no meaningful product distinctions, it gives shoppers and product-matching systems less useful information for deciding whether this item fits a request.
Review the product description and attributes for details that are true, specific, and useful for choosing between alternatives:
- What is the item designed to do?
- Which use cases is it suitable for, and which are outside its intended use?
- What constraints does it meet, such as dimensions, weight, capacity, material, compatibility, or environment?
- What makes this version different from other products in the same catalog?
- Which facts can be verified from specifications, documentation, or genuine customer evidence?
Avoid stuffing a description with every possible search phrase. A long paragraph can still be low-information if it does not clarify product fit. Equally, do not invent specifications, use cases, certifications, or benefits merely to make a product sound more distinctive.
5. Make shopping context explicit
AI shopping questions are often expressed as situations and constraints rather than as a product's exact name. A shopper might ask for a lightweight carry-on for a strict airline size limit, a skincare product for sensitive skin, or a desk lamp for a small home office.
The product data should make relevant facts explicit where they are true and applicable. Depending on the category, useful context can include:
- Intended user or audience
- Dimensions, weight, capacity, and materials
- Compatibility and supported standards
- Conditions of use and limitations
- Included accessories and what is not included
- Care requirements, fit, installation, or setup
- Price, availability, and variant-specific differences
Use fields and structured attributes when the platform supports them, then make the most important details visible in normal page content as well. Structured data is not a substitute for accurate product information, and a claim hidden only in an inaccessible script is harder to verify than a clear, visible fact.
For products with meaningful variants, make it easy to distinguish the parent product from each selectable version. Google documents the ProductGroup, hasVariant, variesBy, and productGroupID properties for describing product variants in Google Search. Those properties are a documented structured-data approach for Google, not a published guarantee of how ChatGPT ranks shopping results. See Google's product variant documentation.
6. Treat brand credibility as evidence, not a secret ranking score
Brand information can help make a product's identity and source clearer. Check that the brand name is consistent across the store, product records, business profile, and legitimate external profiles. Make sure the official store, manufacturer information, and product claims are not contradictory.
Useful checks include:
- A clear About or brand page that explains who operates the store
- Accurate brand and manufacturer fields in product data
- Working links to official social and business profiles, where relevant
- Current shipping, returns, and contact information
- Genuine reviews or documented product experiences, presented honestly
- Consistent facts across product pages, collections, policies, and structured data
ChatGPT's public shopping explanation says merchant listings can use factors such as availability, price, quality, and whether the merchant is the maker or primary seller. It does not publish exact weights for these factors, nor a universal "brand authority score." Do not tell clients that adding an Organization schema or sameAs links will automatically increase product recommendations. These are identity and consistency checks, not ranking guarantees. See OpenAI's explanation of shopping search.
7. Separate web crawling from Shopify Catalog distribution
Crawler checks matter, but they answer a different question from catalog eligibility.
OpenAI documents separate agents for different purposes:
- OAI-SearchBot is used for ChatGPT search discovery.
- GPTBot is used for crawling content that may be used to train OpenAI foundation models.
- ChatGPT-User can fetch a page when a user explicitly asks ChatGPT to visit or inspect it; it is not the automatic search crawler.
These controls are independent. Allowing GPTBot does not guarantee a product listing, and blocking GPTBot is not the same as disabling a Shopify Catalog feed. If your problem is a website page missing from ChatGPT web-search results, review the relevant search crawler rules. If your problem is a shopping product card missing from a Shopify-connected catalog, first check the catalog channel and product data.
OpenAI's current reference is Overview of OpenAI Crawlers. Google also clarifies that Google-Extended is a robots.txt control token, not a separate HTTP User-Agent string, and does not affect inclusion in Google Search. See Google's crawler reference. Apple likewise describes Applebot-Extended as a control for use of content, not a crawler that fetches webpages itself: About Applebot.
If you control your site's robots.txt, CDN, WAF, or reverse proxy, check the relevant rules and logs. If you use standard Shopify hosting, you may not have direct control over every edge-security layer. Do not blindly copy a universal list of AI bots into your configuration; decide which access you want to allow, then use each provider's current documentation.
For web-crawl paths, inspect:
- Whether relevant product URLs are disallowed by robots.txt
- Whether broad query-parameter or filter rules accidentally affect canonical product URLs
- Whether a CDN or WAF returns 403, 429, or a challenge page to legitimate crawlers
- Whether the initial HTML contains substantial product information or depends on client-side rendering
- Whether canonical tags match the intended indexing strategy
- Whether the sitemap contains the canonical product URLs that should be crawled
Do not assume every product variant needs its own sitemap URL. The correct setup depends on whether variants have distinct, crawlable URLs that represent distinct pages. Align the canonical strategy and structured data with the actual page architecture.
8. Test recommendations with controlled prompts
After technical checks, run a small, repeatable test set rather than asking ChatGPT one broad question and treating the answer as a ranking report.
Start with prompts in four groups:
- Category: "Recommend wireless headphones under $200 with multipoint Bluetooth."
- Comparison: "Compare Brand A with Brand B for commuting and frequent video calls."
- Problem or use case: "What products are suitable for someone who needs a compact, waterproof daypack for cycling?"
- Brand-specific: "What should I know about Brand X and its Model Y?"
For each test, record the exact prompt, date, market, model or surface, whether web search was enabled, whether your product was included, whether your URL was cited, which competitors appeared, and how the product was described.
For a first diagnostic pass, 20 to 30 well-designed prompts may be enough to reveal obvious patterns. A more systematic audit can use 30 to 100 prompts, repeated across more than one provider or session. The larger count is a sampling choice, not a universally validated threshold.
Do not mix browsing-enabled answers with non-browsing answers in the same score. They answer different questions. Repeat the same prompt set over time because results can vary by location, query wording, model version, shopping inventory, and other contextual factors.
9. A practical troubleshooting order
Use this order to avoid optimizing the wrong layer:
- Define the symptom: no channel access, incorrect product data, no result for one query, or no measurable visits.
- Check Shopify Catalog eligibility and channel settings.
- Verify the product data and mappings that the channel actually receives.
- Review product-specific details: variants, price, availability, compatibility, use cases, and meaningful differences.
- Review web crawling separately if your issue concerns search discovery or the public product page.
- Run repeated, scenario-based recommendation tests.
- Measure visits and orders separately. A product mention does not prove that a customer clicked, and a click does not prove incremental revenue.
What this diagnosis can and cannot tell you
This process can uncover access problems, mismapped fields, unclear product information, and patterns in a defined sample of AI shopping answers. It cannot reveal a platform's private ranking weights, guarantee a recommendation, prove why an individual answer omitted a product, or promise that a change will increase sales.
The useful output is a short list of evidenced issues, the products affected, the recommended correction, and a plan to verify the outcome. That is more actionable than a generic claim that a store is or is not "AI ready."