AI shopping turns product data into a commercial control
A store can have the right product and still be excluded or represented incorrectly when its website, catalogue and feeds disagree. This is no longer a narrow advertising-feed problem.
On 10 November 2026, OpenAI introduced Businesses in ChatGPT in beta, including verified business information and self-service product feeds. Google says accurate product data supports matching and AI-powered experiences; eligible free listings can appear across Search, Shopping, Images, Lens, YouTube and Gemini. Neither platform guarantees an appearance.
My verdict is direct: treat product data as governed inventory, not channel copy. A new feed or schema plugin is useful only after leadership can state which system owns product identity, variants, price, stock, images and fulfilment promises.
The ThomPerformance Commerce Discovery Truth Contract
The contract is a five-part owner check. It converts a technical catalogue into business promises that can be maintained and reconciled.
Know what the product is
Use stable IDs, accurate brand and manufacturer identifiers, a factual title, a useful description and a destination for the exact product.
Keep options distinct
Represent each purchasable size, colour, material or configuration correctly while preserving its relationship to the parent product.
Match price and availability
The feed, structured data, product page and checkout must agree on the current price, currency, stock state and applicable offer conditions.
Make the promise purchasable
Shipping, delivery geography, return policy and seller identity should help a buyer judge whether the product can meet the intended need.
Detect drift before customers do
Assign an owner, refresh cadence, exception route and a regular comparison across the commerce platform, feeds and live product pages.
Google warns that missing identifiers, variant attributes, images or conflicting website data can cause disapprovals, limited eligibility or incorrect display. OpenAI likewise calls for current titles, descriptions, images, price, availability and variant-level accuracy. The platforms use different schemas; the company truth underneath them should not differ.
Use the Product Truth-to-Purchase Evidence Loop
A clean feed is not the finish line. Connect catalogue accuracy to discovery, suitable visits, purchases and retained contribution.
Set the source
Name the authoritative system and business owner for each material product fact.
Distribute consistently
Send the same current truth to product pages, structured data, Merchant Center and eligible AI-shopping integrations.
Resolve exceptions
Separate source-data errors, transformation errors, policy issues and destination-page conflicts.
Measure suitable discovery
Track eligible products, impressions or citations, product visits and the queries or contexts that expose the catalogue.
Follow the order
Connect product discovery with purchases, margin, cancellations, returns and new-customer value.
Fix the weakest truth
Prioritise the product or attribute creating the largest commercial loss—not the longest diagnostics list.
ChatGPT says shopping results may consider structured metadata and warns that updates can take time. Monitor appearances, but do not claim a stable AI ranking or attribute every later order to one product appearance.
Choose the next investment by the evidence you have
| Current evidence | Decision | First commercial action | Do not do this |
|---|---|---|---|
| Store data is wrong or incomplete | Repair the source | Correct identity, variants, price, stock and policy ownership in the commerce system | Copy different fixes into every channel |
| Source is accurate; feeds disagree | Repair distribution | Audit mappings, refresh timing, currency, markets and platform diagnostics | Rewrite titles before fixing stale price or availability |
| Products are eligible; suitable discovery is weak | Improve relevance | Strengthen factual titles, descriptions, categorisation, identifiers and product evidence | Stuff generic keywords or invent attributes |
| Discovery grows; purchases do not | Repair the buying path | Compare product promise, price, shipping, reviews, mobile experience and checkout | Add more feeds and call reach the solution |
| Purchases grow; contribution falls | Protect economics | Reconcile product margin, discounts, media, fulfilment, returns and customer mix | Scale from platform revenue or feed clicks alone |
For adjacent decisions, review marketplace versus owned store, whether TikTok Shop fits and which product to promote first.
Illustrative example: one product, three different promises
Illustrative example—not a client result: a lamp is in stock at £120 with three-day delivery. One feed shows £105 and both finishes available; structured data reports the wrong variant.
More reach would multiply the contradiction. The business assigns its commerce platform as the source, corrects theme output and feed mapping, shortens the refresh window, then checks each destination. Its scorecard separates eligibility, suitable appearances, visits, purchases and contribution after returns. Feed approval is operational evidence—not proof of profitable demand.
A 30-day owner plan for shopping-ready product data
Define truth
Select priority products and name the source, owner and acceptable update delay for each critical fact.
Compare surfaces
Check the store, structured data, Merchant Center and eligible AI-shopping feed for material disagreements.
Repair one chain
Fix source data, mappings and refresh logic for one commercially important product family before scaling the pattern.
Reconcile outcomes
Confirm live accuracy, then observe suitable discovery, product visits, purchases and contribution with explicit attribution limits.
This is a control cycle, not a 30-day visibility or revenue promise. Some integrations require eligibility or partner access, and discovery systems decide what to surface.
Connect it to Google Ads, AI Growth, conversion tracking and profitable ecommerce scaling. Review approved case evidence, evidence standards and Thomas's operator model.
Sources and evidence notes
Official sources were checked on 11 November 2026. The frameworks are original ThomPerformance analysis; the example is illustrative.
- OpenAI: Introducing Businesses in ChatGPT (10 November 2026)
- OpenAI Developers: getting started with Agentic Commerce product feeds
- OpenAI Help Center: shopping with ChatGPT Search
- Google Merchant Center: product data specification
- Google Merchant Center: free listings for products
- Google Search Central: merchant listing structured data
Frequently asked questions
What product data matters most for AI shopping?
Start with the facts that decide whether a product is suitable and purchasable: a stable identifier, clear title and description, correct variant, current price and availability, representative images, destination URL, seller identity, shipping and return information. The exact required fields vary by platform, but accuracy and agreement with the store are the commercial foundation.
Do Shopify merchants need a separate ChatGPT product feed?
OpenAI says eligible Shopify products are already integrated through Shopify Catalog, and its shopping help page says individual Shopify merchants do not need additional work for that integration. Merchants should still verify product, price, stock, image and policy accuracy in Shopify because an integration cannot repair incorrect source data.
Is product structured data enough for Google Shopping visibility?
No. Product structured data can make a purchasable product page eligible for merchant listing experiences, but eligibility does not guarantee visibility. Google also uses Merchant Center data, website content, policy compliance and product-page consistency. Use structured data and a maintained product feed as complementary evidence, not competing shortcuts.
Will a better product feed guarantee more sales?
No. Better data can improve eligibility, matching, accuracy and buyer confidence, but it cannot guarantee placement or sales. Product demand, price, merchandising, reviews, shipping, conversion experience, competition and customer fit still matter. Measure qualified product discovery through to contribution after returns rather than treating feed approval as revenue proof.
Who should own product data quality?
A commercial owner should be accountable, even when ecommerce, merchandising, operations, developers and agencies maintain different systems. Ownership means agreeing the source of truth, update cadence, exception route and business thresholds. A channel specialist can diagnose feed errors, but should not silently decide product identity, price, stock or fulfilment policy.
Make product truth portable before making reach larger
Govern the source, model the purchasable variant, keep price and stock current, clarify fulfilment and reconcile discovery with retained economics. Then expand the channel—not before.
