ResourcesGuide
Wrong AI answer: Shopify Knowledge Base or product data?
Route a disputed store-policy or product answer to the source your team owns, then verify the exact answer in the route that matters.
Updated 3 min read
When an AI shopper gives a wrong answer, first ask what kind of fact was wrong. “Can I return a used item?” and “Does this adapter fit Model X?” need different owners and sources. Editing a store FAQ will not repair a wrong variant record; rewriting a product title is a poor substitute for an up-to-date return policy.
This guide helps you choose the first native correction. It does not assume that you can see every source an outside AI channel used.
Route the claim to its source
Scroll sideways for columns out of view. Keyboard: focus the table and use the Left and Right arrow keys.
| Contested claim | First place to inspect | Why |
|---|---|---|
| Return window, shipping service, store hours | Published policy/service content and Shopify Knowledge Base facts or FAQs | Shopify positions Knowledge Base for store information and FAQs. Check for a conflict between the policy and answer. |
| Product size, material, specification or model fit | Maintained product/variant record, linked specification or compatibility source | Establish the exact item and reference before changing any copy. |
| Product title, description, options or grouping shown in AI discovery | Shopify Catalog data and applicable Catalog Mapping | Mapping lets merchants choose sources for documented product fields and grouping. |
| Stock, price or delivery for a selected option | Exact variant, market and time in Shopify’s current commerce data | A later snapshot cannot prove what an earlier answer saw. |
| “Can I return this specific configured item?” | Both policy and product/variant exception | Mixed claims may need two owners; do not flatten them into one FAQ. |
These are starting points, not proof of retrieval. Keep the answer and source in the answer log; mark the source unknown if the session does not expose it.
Use Knowledge Base for store facts
Shopify says its Knowledge Base app lets merchants preview generated store facts, review common questions, and customize or create FAQ answers. Shopify also says the app does not change how often a store appears in AI results. Review the current policy first, then correct an outdated or inconsistent FAQ. Shopify Knowledge Base.
Illustrative example: a return FAQ says 14 days while the published policy says 30 days for the item in question. Ask the policy owner which term applies, correct the authoritative policy/FAQ conflict, and retain the before and after records. Do not infer from a later AI answer that this FAQ alone caused the change. A channel may have used another source.
If the question is “Does this exact item fit Model X?”, a broad store FAQ may hide the relevant variant or revision. Use a maintained product relationship and test the exact identity instead.
Use the product owner for product claims
For a specification, variant or compatibility claim, ask the product-data owner to review the exact item. Shopify Catalog and Catalog Mapping are native sources to inspect for connected AI channels. Inclusion does not promise placement or wording.
Use the product-data trace to follow the fact into a page, channel or browser response. A broad store FAQ should not replace a model-specific relationship.
Verify the correction with its owner
Next step: assign the contested claim to the source owner in the table above. Resolve any policy/product exception together, then use the comparison method to repeat the question on the relevant route. An altered answer is a new observation, not proof one FAQ edit caused it.
StoreKnows’s simulated product checks and supported website tools serve a compatible visiting-agent route. They do not replace Knowledge Base, rewrite Catalog feeds or validate product fit through a store FAQ.
StoreKnows is independently developed and not affiliated with, endorsed by or sponsored by Shopify or the AI platforms discussed.
Investigate the answer at its source
Once you know whether the question concerns store policies or product facts, follow the relevant source instead of repeating the same test everywhere.
Diagnose a product-answer gap