This is a complete sample of the Shopify agent-readiness audit report, written for a fictional store so you can judge the depth, format and tone before ordering. The real report contains only facts from your store.
Read this first: "Example Store" is a fictional mid-size home-goods store created for this page. It is not a client, and no real store's data appears here. The findings below are composites of the issues the 12-point check looks for, at the level of detail a real report carries.
| Store | Example Store (fictional) · mid-size home-goods retailer, ~140 products, standard Online Store 2.0 theme |
| Audited | Illustrative example · findings are composites, not from a real engagement |
| Scope | Public storefront (home, collections, 5 product pages), /agents.md, /llms.txt, structured data on 3 product URLs, a full checkout walk-through in a private window, policy pages |
| Delivery | Written report (PDF) by email, within 5 business days of order |
| Result | 5 pass · 3 partial · 4 fail of 12 points — agent-readable, not yet agent-reliable |
Example Store is agent-readable but not yet agent-reliable. The good news: products are discoverable, structured data is valid, and checkout is clean - an AI shopping agent can find a product and complete a purchase today. The blockers are in the facts layer: the Knowledge Base has never been published, so policy questions get no store-specific answer at all, and /agents.md states a 30-day returns window that contradicts the store's real 14-day policy. Product specificity is too thin for agents to match the questions shoppers actually ask - "Option 1" variants and brand-only titles lose every specific match to a competitor. The pre-Black-Friday list is short and cheap: publish the facts, fix one 404, and enrich the twenty products that matter most.
Every finding states what we saw, why it matters for agents, and the exact fix with a rough effort estimate. In a real report each finding also links the exact URLs and includes screenshots of your store.
Finding: the Knowledge Base (Sales channels > Agentic > Knowledge base) still holds drafted facts; nothing has been published. The returns-window field is blank and shipping regions list only the domestic market.
Why it matters: unpublished facts are invisible to agents. Until you press Publish, agents have no store-specific answers for returns, shipping or gift questions.
The fix: fill in the returns window (14 days), shipping regions and times, and gift options, then publish.~30 min
Finding: /agents.md loads and is well-formed, but it states a 30-day returns window while the published refund-policy page says 14 days.
Why it matters: whichever the agent quotes, the answer is wrong for some shoppers - and a wrong returns promise at purchase time turns into a dispute at return time.
The fix: correct the fact in the Knowledge Base and add a templates/agents.md.liquid theme template so the file states the real policy; re-check the URL after publishing.~1 h
Finding: /llms.txt returns 200 and mirrors /agents.md.
Why it matters: agents that request /llms.txt instead of /agents.md still get a usable document.
The fix: none needed now - re-verify after any /agents.md change.
Finding: 118 of 140 titles are brand plus a generic noun ("ExampleCast Premium Carafe"); capacity, material and compatibility never appear in the title.
Why it matters: agents match titles against specific questions ("a 600 ml glass pour-over carafe"). A generic title loses that match to whoever names the specifics.
The fix: rename the 20 best sellers first: material + size or capacity + compatibility in the title.~3 h for top 20
Finding: 2 of 5 tested product pages state dimensions, material and care; the rest carry a single line plus "see photos".
Why it matters: the description is what an agent quotes. "See photos" gives it nothing to answer with.
The fix: 3-4 plain sentences per product: dimensions, material, care, compatibility. Start with the best sellers.~15 min per product
Finding: Product and Offer schema present and valid on all 3 product URLs tested (Google Rich Results test).
Why it matters: price and availability reach agents in machine-readable form, not just rendered text.
The fix: none. Re-test after any theme change.
Finding: checkout completed in a private window without account creation.
Why it matters: forced account creation is where agent-led purchases stall and die.
The fix: none.
Finding: no bot walls, app-install screens or password gates between product page and payment.
Why it matters: agents cannot pass interactive walls; the purchase would simply fail.
The fix: none.
Finding: refund and terms pages load and are footer-linked; the shipping-policy link 404s on both desktop and mobile.
Why it matters: "how long does shipping take?" is a top agent question, and a 404 removes your answer while looking broken to trust signals.
The fix: recreate the page under Settings > Policies and relink it in the footer.~30 min
Finding: Shop product reviews render on the tested product pages, not locked inside an app popup.
Why it matters: visible reviews are facts an agent can weigh ("highly rated for sizing true to fit").
The fix: none.
Finding: 9 sold-out variants still show an active Add to cart button, with no restock signal or notification option.
Why it matters: an agent that recommends an unavailable product loses the shopper's trust for the next question.
The fix: mark items sold out and add a notify-me option.~1 h
Finding: every product uses "Option 1" and "Option 2" instead of Size, Color and Material.
Why it matters: "Option 1: Large" cannot answer "does it come in green?" - the option name carries the meaning.
The fix: rename the options in admin; no data migration involved.~2 h across the catalog
Every real report ends with this list: what to fix before Black Friday, what can wait, and what each fix is worth in agent terms.
The audit covers your live store: storefront, /agents.md, /llms.txt, product data, structured data, checkout walk-through and policies. Fixed price, fixed scope.
Fixed scope agreed in writing before any work starts. Card payment via the links above; EU businesses can request an invoice instead - just email us. Prefer to self-check first? The free 12-point self-check stays free.