What an AI Shopping Assistant Needs to Know About Your Products

Two cartridges can look identical in a product photo and still fit different printers, so a shopper who types “black ink for a home printer” has supplied the color and left out the fact needed to make a match. The printer’s model number usually appears on a label the buyer can check. Until the buyer supplies it, a confident recommendation risks sending the wrong cartridge to the right address.
Commerce introduced Shopping Agent for BigCommerce on November 16, 2026, and began offering it to eligible merchants. Shoppers can compare products on the merchant’s site and carry details across follow-up questions. Commerce separately said its conversational storefront search was available on all paid BigCommerce plans. Access to storefront search and eligibility for the shopper-facing assistant remain separate. Merchants in Dubai should confirm availability for their own store; the announcement describes expansion to further regions over time and leaves country dates and product pricing open.
Identify the Product the Shopper Can Use
For a shop selling printer supplies, the product family helps people browse, but the cartridge itself is the item the merchant can recommend. Store the printer manufacturer and supported model designation against the cartridge record. Where cartridge compatibility varies by region for the same printer model, retain the region in the compatibility record. Even a search result for the correct brand requires an exact model check.
Product names help the assistant find candidates when buyers describe a need in ordinary terms, such as asking for ink for a small home printer. Keep common aliases alongside verified model names so the system can recognize how shoppers type. For compatibility, use the manufacturer’s supported-model table or the merchant’s verified product records. If a supplier offers a replacement cartridge, record which models the supplier lists and who will resolve a disagreement.
Apparel merchants face a related issue when a customer asks for a size that appears on several brands’ labels. One maker’s medium may measure differently from another’s, and the relevant size chart belongs with the actual garment. Record the variant’s measurements and keep the measuring method beside them. For a jacket, the chart may need to distinguish the buyer’s chest measurement from the garment’s finished width.

Separate Requirements from Preferences
If a buyer asks for a replacement cartridge for a specific printer and sets a delivered limit of AED 160 in Dubai, both the physical match and the stated ceiling determine eligibility. The delivered price uses the store’s current offer and the applicable delivery charge. If checkout calculates the charge based on an address, the assistant can state the item price and request the minimum location details needed for a delivered quote.
Once the store identifies cartridge offers that meet the customer’s requirements, it can rank them using the buyer’s stated preferences. If both the original cartridge and a third-party version qualify, explain the difference using facts the merchant can verify. The lower-priced option warrants an explanation for its appearance, such as a confirmed fit at a lower delivered cost. Only a customer’s revised budget makes a dearer alternative eligible.
Some shoppers provide a cartridge code instead of a printer model, so the shortest useful question depends on how the shop records the relationship. If the old cartridge identifier maps to one compatible replacement, the system can check it against the reference table. When one description maps to several incompatible variants, ask for the exact code printed on the old cartridge or the printer’s model label. Keep the shopper’s earlier budget and delivery requirements attached to the search.

Keep Specifications and Live Offers Apart
Merchants typically revise fit information when a product maker updates its specifications or a catalog team finds an error; stock and price can change during one conversation. The assistant can explain compatibility from the specification record. Check the applicable storefront’s price and inventory when the assistant prepares an offer for the buyer. If the inventory reading is stale, query it again before presenting the recommendation.
When a policy applies only to certain products or locations, connect delivery and return terms to the item and market the shopper selected. If arrival depends on the district, request the information needed to check the date. Sealed and opened versions of a product may carry different return terms. Where the general returns page conflicts with the offer, route the question to the team responsible for the applicable policy.
Google Merchant Center’s variant guidance provides a separate example of how a product feed can preserve family relationships while identifying individual offers. Variants share an item_group_id, and the details for the selected item should match those on the landing page. Google specifies these rules for its own feed. Merchants can apply the comparison to their own catalog: verify the displayed product against the record the assistant chose.

Check the Offer Once It Leaves the Conversation
When a buyer opens the recommendation for a compatible cartridge, the page should show the agreed variant and the current price for the buyer’s market. The Google Search documentation on variants describes direct links that preselect a variant and display its own offer information; shoppers can add the selected variant to the cart. Test the handoff by comparing the recommendation with the product page. Add the chosen item and confirm the cart carries the same identifier and applicable offer.
BigCommerce’s Shopping Agent product page describes adding items to a cart and directing the buyer to the store’s checkout, where payment occurs outside the conversation. In a separate November 19 developer update, BigCommerce made Store Management MCP available in early access for all stores. The new access applies to merchant administration; users receive permissions from an administrator. Treat the two announcements as separate: shopper access follows the Shopping Agent rollout, and store administration follows its own access rules.

Measure the Quality of the Product Decision
Retailers can build a test set from the questions customers already ask concerning product fit and the mistakes that have led to returns. Use product questions with independently checked answers. Include cases where the expected result is a follow-up question or a referral to support, such as an unverified fit or a return term that requires confirmation. Save the expected decision with the relevant catalog evidence so a second reviewer can check the result.
When two tests fail for different reasons, send the evidence to different teams so the fix reaches the right part of the store. Say the test cartridge fits and costs AED 150 delivered, but the assistant recalls an earlier price of AED 165 and drops it from the list. Check the offer lookup and repeat the test with the current price. If the price was right and the wrong cartridge made the list, review the compatibility entry or the rule that decides which models qualify.
Retailers measuring recommendation quality can record the share of responses that is supported by product evidence. Count a request for a missing model number as the correct result when the request reaches a decision the store needs the shopper to make. Check exact-fit recommendations separately from cases where the system paused to ask for essential information. Track whether offers were available at the time they appeared. Record confirmed incompatibility separately from a change of mind when shoppers return items. Measure any claimed effect on conversion through a separate comparison suited to the store’s traffic and product mix.
For a storefront engineering review, start with a category where product differences govern purchase decisions, such as printer compatibility or garment size. Use questions the support team already answers by hand. Where the store has a trusted source for the answer, assign it to a specific record and test the match. Send unresolved product questions back to the category owner; ask the team responsible for terms to resolve any policy discrepancy.
Make your brand matter.

