ResourcesGuide
Your tasting notes already exist. Check whether AI shoppers can use them.
Check whether a storefront agent receives your published coffee facts. StoreKnows’s baseline does not ask tasting-note questions; a supported proposed-fix preview can let you try your own.
Updated 4 min read
If a shopper asks how two coffees differ, your published origin, process and tasting notes should be the reference. Start by checking whether the visiting agent receives those facts through the product page or Shopify’s native tools.
StoreKnows’s free baseline does not ask tasting-note questions. It builds questions from fixed catalog templates. There is no dedicated tasting-note tool. Published text fields can appear in supported filter or comparison proposals; where a proposal exists, Try a shopper question lets you test your own question with that fix and a saved copy of your catalog. Review a proposal’s actual field coverage before deciding to publish it.
This is a conditional use case for a merchant who maintains the deciding facts, not evidence that coffee stores need an additional tool.
Inspect where the facts appear
Shopify’s native website tools can return product details. Shopify also describes agents reading page code and interacting without those tools. The useful question is what this session received, rather than whether a format is always readable.
Scroll sideways for columns out of view. Keyboard: focus the table and use the Left and Right arrow keys.
| Where you publish the fact | What to check |
|---|---|
| Product description or tasting-note text | Does the response include the exact notes for the correct coffee and lot? Does another field contradict them? |
| Bag image, roast card or PDF | Did this agent receive and read it? Vision-capable models can interpret images, but small or rotated text can cause errors. |
| Accordion, tab or JavaScript panel | Can the agent expand or load it? Is the text present in the retrieved page content? |
| Metafield or coffee-profile metaobject | Does it appear on the page or in the tool response? Its presence in admin alone does not settle that question. |
Structured, labelled text can make these facts easier to retrieve and review. A collapsed panel or JPG is a place to investigate, not proof that every agent misses the information. If the native tools and page already answer well, extra publication may add no useful value.
An illustrative comparison to review
The following is a fictional example of source facts and a possible answer, not a StoreKnows run, proposed tool capture or customer result.
Scroll sideways for columns out of view. Keyboard: focus the table and use the Left and Right arrow keys.
| Saved product fact | Coffee A | Coffee B |
|---|---|---|
| Origin | Colombia | Ethiopia |
| Process | Washed | Natural |
| Published tasting notes | Cocoa, orange | Blueberry, jasmine |
Question: “Compare the published origin, process and tasting notes of Coffee A and Coffee B.”
Illustrative supported answer: “Coffee A lists Colombia, washed processing, and cocoa and orange notes. Coffee B lists Ethiopia, natural processing, and blueberry and jasmine notes.”
A comparison proposal is relevant only if it actually exposes those fields for both products. Inspect the saved data and try the question in the proposal’s preview. If only origin and process are available, the answer should acknowledge that tasting notes are unavailable. It should not infer a cupping score, brew recipe or which coffee you will prefer.
A plausible answer is insufficient evidence of improvement. Compare it with the native answer using the same products and question; keep correct native answers, unchanged results and unsupported claims in your record. See how to judge before-and-after answers.
Prepare one useful question
Keep product and lot identities clear, use your published values, and resolve contradictions between descriptions, panels and metafields. Record the source date; published tools use saved data, so a new lot or changed tasting profile needs review and an updated publication.
Then:
- Choose one comparison that matters to a buyer and the facts you would use to answer it.
- Save the native question and answer, plus the page or tool response where available. A readiness score may help prioritize source work; it cannot tell you whether this particular answer is supported.
- If a supported proposal exists, inspect its fields and try your question in Try a shopper question. One model answers using that fix and saved product data; this preview is separate from the catalog-built baseline.
- Review the answer against your facts. Defer publication if data is missing or the preview adds no useful information.
This investigates a compatible agent using tools on your storefront. It does not establish changes to Shopify Catalog, AI discovery or consumer-assistant recommendations. For the route distinctions, see which product data agents can read.
Check the catalog, then evaluate an applicable proposal
See how the free check works to decide whether its question coverage fits your catalog. Install StoreKnows for one free catalog-built check per store. If an applicable fix appears, try your own question before deciding what to publish.
Updated 2 October 2026: moved the baseline limitation to the opening, qualified page and image reading, and replaced hypothetical before-and-after transcripts with one labelled comparison illustration. Platform reading documentation was checked on 2 October 2026; product scope uses the retained 29 September app review.
StoreKnows is independently developed and not affiliated with, endorsed by or sponsored by Shopify, OpenAI, Google or Anthropic.
Choose a question the check can cover
Use the question playbook to separate your tasting-note investigation from supported product-question tests.
Choose and record a product questionQuestions this article answers
- Does the StoreKnows free check test tasting-note questions?
- No. The baseline builds questions from fixed catalog templates covering specifications, compatibility, variant availability and comparisons. Where a supported fix is proposed, Try a shopper question lets you ask your own question using that fix and saved product data. There is no dedicated tasting-note tool.
- Can an AI shopper read a bag photo or collapsed tasting panel?
- It depends on the agent, the page and what the session retrieves. Some agents can interact with pages or use vision. Inspect the actual content and answer before calling a format unreadable. Clear labelled text can make the facts easier to check.
- Will StoreKnows supply tasting notes or scores we have not published?
- The proposed tools expose supported saved store facts; they do not supply unpublished tasting notes or cupping scores. Review the model’s answer against those facts and treat unsupported claims as errors.
- What should I test first?
- Choose one product distinction you maintain, such as the published origin and process of two named coffees. Inspect the native response first. If a supported comparison proposal includes those fields, try the same question in its preview and review the answer.