KINDERAI / MERCHANT GUIDES
Test product recommendations in your Shopify chatbot
A product recommendation should respect what the shopper asked for and lead to an item they can actually evaluate. Test the complete journey from question to product to cart, using a small selection from your own catalog.
Choose a useful test collection
Include an available product, one with several sizes or colors, a sold-out item and a product outside the test budget. Confirm the current storefront prices, descriptions and variant availability before you begin.
Ask for more than a category
Try a concrete request such as “a linen shirt under $70 in medium.” Check each constraint separately. If nothing matches, a helpful assistant should explain that limitation rather than quietly recommend a different material or a higher price.
Follow up naturally
Ask “Does it come in blue?” or “Is there a cheaper option?” Check whether the assistant keeps the relevant product context and whether each new suggestion matches the updated request. Then start a new conversation and make sure an old preference does not create an irrelevant answer.
Inspect the product and variant
Open the product page. Check title, currency, price, material and availability. Where a product has multiple variants, verify the chosen combination. An attractive card is not evidence that the exact size can be ordered.
Check the actual cart
Add an available variant and compare the item and quantity in Shopify’s cart. Repeat with an unavailable option. Watch for duplicates and confirm the assistant does not report success when nothing was added. You do not need to pay or place an order for this check.
In KinderAI
Select the products the assistant should use in Settings and save. Use active products published to the Online Store, and confirm available variants for cart tests. The website preview uses sample products and a simulated cart; it does not sync your catalog or prove live-store behavior.
Try the interactive preview ↗Measure useful outcomes first
Record matching recommendations, failed requests and correct cart actions. Review where shoppers need more information. Treat sales after a chat as an observation until you have an attribution method that supports a stronger claim.
Continue with the full evaluation checklist, store FAQ preparation or setup troubleshooting.