Edu here. This week I was reading Shopify's update about the facts it gives AI shopping agents, and I kept thinking about the questions people ask before buying. Will it leak? And can I return it after opening the box, even if I've already used it? A page can sound convincing and still leave someone guessing.
On October 26, 2026, Shopify moved Knowledge base into its admin, under Sales channels > Agentic. I went through the documentation with a small brand's unanswered questions in mind.
Key Takeaways
- Shopify's October 26, 2026 update brings Knowledge base into the admin.
- Catalog supplies product information; Knowledge base supplies store facts.
- Check which fields Catalog receives, especially when descriptions live in custom fields.
- Use customer questions to find missing facts, then confirm the answers before publishing them.
What changed in Shopify Agentic Storefronts this week?
In its October 26, 2026 announcement, Shopify's “Knowledge base is now built into your admin” says the standalone app will close after January 15, 2027. The new admin page lets merchants review and publish store facts, with links back to the storefront pages those facts came from.
Publication depends on the source, according to Shopify's Knowledge base help page. Storefront scan suggestions need approval, while admin-linked policies and contact details are published when created. Existing app users have separate migration rules (check each fact's status).
I expected another announcement about generating answers, but the source links caught my attention. You can check them. Maybe your shipping promise sounds clear, until someone asks whether the quoted time includes packing. Follow the link and see whether the original page actually says.
Product data in Catalog and store facts in Knowledge base
As of October 2026, Shopify's “Optimizing your products for AI platforms” lists product titles and descriptions among the fields to improve. It directs merchants to Knowledge base for non-product information. Treat them as separate places to check, because a good returns answer doesn't fill in a missing product detail.
| Customer question | Where I'd check the answer |
|---|---|
| Will this bottle fit my cup holder? | Product dimensions and the relevant variant's details in product data. |
| Does the lid seal when I put it in a bag? | The product description, backed by the brand's actual testing. |
| Can I return it after opening the box? | The returns policy and its published Knowledge base facts. |
| Can you gift wrap it? | The store's gift services information in Knowledge base. |
In this hypothetical example, “made for everyday adventures” tells me little about a bottle's suitability. Dimensions and lid instructions help, so I'd look for missing details in customers' own questions (before choosing new copy).
Questions identify concerns, but they don't prove answers. Ask whoever tests the bottle what happened when they carried it sideways. I think customer research earns its place when it leads to a specific fact someone can confirm.
Does Shopify receive the description your customers see?
In October 2026, Shopify's “Customizing your product data for Shopify Catalog” describes mapping the product title and description to their source fields. It specifically gives the example of a full description stored in a metafield. Check that mapping if your storefront uses custom content, because the visible page and supplied description can differ.
A metafield holds extra information about a product, and maybe your developer stored the full description there while leaving a short standard version. Compare both in Review product data mapping (especially with custom layouts).
The preview shows Catalog's input, and Shopify warns that the final AI display can differ. Inspect that separately. I don't know how much an edit improves placement, because we haven't run that experiment.
In our 2026 AI shopping recommendations study, about 51% of coded complaints about recommendations concerned stale or wrong information. The sample comes from people discussing AI shopping online, so the percentage describes complaint mentions (not a rate among all shoppers). One commenter put it plainly:
“Don't use Chat GPT or any other AI service as a search engine. They don't actually "know" anything and if the data they're trained on contains inaccurate information they'll still repeat it as if it was true.”
How do customer questions become better product information?
Collect recurring buying questions, confirm the answers with your team, and put each fact in the appropriate product field or store policy. Better information can't prevent every invented answer: roughly 20% of coded recommendation complaints concerned nonexistent products in our 2026 AI shopping study. But shoppers can check your own claims.
I'd compare support messages with category discussions on Reddit or YouTube. What's missing? People unfamiliar with your brand might raise concerns your regular customers have already settled, so keep their original wording (remove personal details from support examples).
| Illustrative question | What the brand should confirm | Useful published detail |
|---|---|---|
| “Will it leak in my bag?” | Which lid was tested, and under what conditions? | The tested use and any limits for each lid. |
| “Is this the smaller bottle?” | Which variant is pictured and selected? | Capacity and dimensions attached to the right option. |
| “Does delivery mean it ships tomorrow?” | When does packing end and transit begin? | Separate processing and carrier estimates in the shipping policy. |
The worksheet uses invented examples, and maybe the real lid leaks sideways. Say so. Clear limits help people avoid unsuitable purchases, even when the copy sounds less exciting. Our landing page research article covers translating those concerns into copy.
Take a hypothetical coffee roaster. A buyer asks for decaf suitable for espresso, but the listing just promises a smooth morning brew. Ask the roaster. What can they confirm about residual caffeine and suitable brewing methods, and which grind options belong beside each variant? Someone ordering whole beans needs a grinder; someone choosing a coarse grind probably has different equipment. Don't infer a certification from a customer's wish for one. This example needs confirmation from the merchant, just as the bottle does.
And keep independent reviews in the picture. In the same research, someone described an AI review that repeated a product page's errors. Product information and evidence from using something serve different purposes:
“it just rehashed and rearranged the spiel on the product page, trouble is the info on the product page was wrong and if they had actually used the product they would have found that out.”
What should a small brand check first?
As of October 2026, Shopify's Catalog documentation says inclusion doesn't guarantee placement in a particular AI answer. Each channel controls its final ranking and wording. I'd use the first pass to check accuracy on a single product, then decide whether a wider cleanup is needed.
Choose a familiar product and collect its unanswered questions. Confirm the answers with whoever makes or ships it, then compare them with Catalog's input and the published store facts. Follow each scanned suggestion's source before approving it. When a policy changes, revisit its associated fact too.
I'm not sure how often this changes a recommendation, and a before-and-after answer won't establish cause. Our earlier article on agentic commerce examines the wider trust problem. I'd do this smaller check because the next shopper deserves accurate information, wherever they ask.
Which questions are your product pages missing?
Insightios researches the conversations where people explain what they need and what makes them hesitate. You get recurring concerns with the original quotes, so you can decide what your pages need to answer.
Explore audience researchFrequently asked questions
What are Shopify Agentic Storefronts?
They connect eligible merchants with supported AI shopping channels, using Catalog for product details and Knowledge base for non-product store information. Eligibility varies. Consult your admin settings to see which channels are available.
Does changing a product description guarantee AI recommendations?
No. Shopify explicitly disclaims guaranteed placement or ranking. Improving a description removes an information gap; it doesn't entitle a merchant to a recommendation or establish a sales increase.
Are Shopify's suggested Knowledge base facts already public?
Scanned suggestions require approval, while admin-linked facts are published when created. Check individual statuses. If you previously used the app, consult Shopify's migration instructions before assuming something remains private.
Where do I check which product description Catalog receives?
Open Sales channels > Agentic, expand the Shopify Catalog source, and select Review product data mapping. The preview helps identify blank fields or incorrect sources, particularly with customized storefronts.
Sources
- Shopify, Knowledge base is now built into your admin, October 26, 2026. Retrieved 2026-10-28. Announcement.
- Shopify Help Center, Answering customer questions with your knowledge base. Retrieved 2026-10-28. Documentation.
- Shopify Help Center, Optimizing your products for AI platforms. Retrieved 2026-10-28. Documentation.
- Shopify Help Center, Customizing your product data for Shopify Catalog. Retrieved 2026-10-28. Documentation.
- Shopify Help Center, Shopify Catalog. Retrieved 2026-10-28. Documentation.
- Insightios, AI shopping recommendations study, 2026. Retrieved 2026-10-28. Study and methodology. Complaint shares and customer quotes above come from this existing study; no new corpus was collected.