Agentic commerce readiness: can an AI shopper find the right product and reach checkout?

Start with one product and one realistic shopping task. Check whether an AI assistant can find it, verify the right variant, compare the total offer and send the buyer to a working checkout.

The practical answer: Start with one product and one realistic shopping task. Check whether an AI assistant can find it, verify the right variant, compare the total offer and send the buyer to a working checkout.

The retailer can lose the sale after the product wins

An AI assistant may recommend your product and still direct the purchase to a different seller. That distinction matters to a brand that sells through its own store and retail partners: product visibility and merchant preference are separate outcomes.

In the LinkedIn post that inspired this guide, Olivia Moore reported a winter-coat search in which a bot block redirected her agent away from Patagonia's site, while REI's availability, pickup and price helped it win. This is her reported observation, not an experiment we ran or a universal judgment about either retailer.

Use the example to ask a better question: if an assistant already wants your product, what stops it from confidently recommending your store?

Know which discovery route you are testing

A web search, a catalog lookup and an assistant operating a browser can encounter different information and different failures. Record which experience you used; do not assume all three share a ranking system.

For Shopify stores, current Catalog guidance explains eligibility, product mapping and catalog distribution. Catalog access and open-web crawler access are separate. Inspect the eligible product count and mapping in your admin before assuming a robots.txt edit will fix a missing catalog item.

Shopify's ChatGPT guidance currently describes a discovery referral that completes the purchase on your online store checkout. Other supported AI channels can have different checkout paths and eligibility. Verify the channel you actually intend to test; do not reuse an old Instant Checkout tutorial without checking its date.

Build a source sheet for one real SKU

Choose a product with commercial importance and manageable complexity. Record its canonical URL, identifiers, exact variant, price and currency, availability, materials or specifications, delivery coverage, shipping charges, return conditions and support route.

Include relevant constraints such as size, compatibility, care, warranty or subscription terms. Make claims specific enough to verify. If local pickup matters, confirm the location, available variant and pickup promise rather than a generic “available nearby” badge.

Compare that sheet with the visible page, the product feed and structured data. A product-level “in stock” claim is misleading if the requested size is unavailable. A sale price without its conditions can create a false comparison.

Google supports product structured data and Merchant Center feeds as complementary ways to supply product information. Its documentation does not turn correct markup into a guaranteed recommendation.

Use prompts that expose buying constraints

Try these with your actual category and constraints. The examples below are hypothetical, not store test results.

1. Find a washable work tote under $120 that fits a 15-inch laptop and ships to West Palm Beach.
2. Compare three options for a lightweight rain jacket, including delivery cost and return conditions.
3. Find this exact product in [size/color] for delivery by [date]. Verify availability on the seller's page.
4. Compare buying [product] directly from the brand with buying it from two retailers.
5. Find an alternative to [product] for someone whose priority is [specific need].

Start with non-branded searches to observe discovery, then use exact-product prompts to test accuracy and merchant choice. Keep those two sets separate. Repeat each prompt in a fresh session, using the same location and requirements, and record the sources.

Ask for evidence supporting stock, price and delivery rather than asking the assistant to reveal an exact hidden ranking formula. Its explanation can be incomplete or mistaken.

Follow the handoff until a human must commit

Open the recommended link. Does it preserve the intended product and variant? Can the shopper read the price, choose an available option and add it to the cart? Check whether discounts, shipping and taxes are handled or clearly qualified before the final commitment.

Look for currency changes, accidental subscriptions, unavailable pickup, a lost variant, an obstructive popup and a cart that quietly substitutes another item. Record the first failure and the screen or URL where it occurs.

For a public audit, stop before submitting payment or an order. Use your platform's approved test mode and an authorized test process when you need to verify order confirmation and fulfillment. A successful cart is not proof of a completed purchase.

Repair the facts and the path together

Correct stale price or inventory data at its source. Improve missing variant attributes and buyer-facing policy explanations. Check whether feed mapping is pulling an outdated field. Reconcile conflicting information across the store instead of adding another AI-written summary.

If a browser agent meets a challenge, have the technical owner examine logs and platform guidance. Keep authentication and fraud controls. On Shopify, network bot management is handled by Shopify; do not add an unsupported proxy or broadly disable protection to make one test pass.

Use accurate reviews and evidence. Do not fabricate customer stories or seed fake forum recommendations. A page that helps a real shopper decide is a more durable asset than copy written to persuade a supposed algorithm.

Keep a product-level scorecard

Track discovery, correct variant, price accuracy, stock accuracy, recommended seller, handoff success and any block. Save the exact prompt, date, channel and evidence URL. Report incomplete tests alongside successful ones.

Make one repair and rerun the same scenarios after the relevant systems have had time to update. Record whether the fact became accurate or the path became usable. Watch returns, support questions and conversion alongside traffic; more visits with worse expectations is not a win.

Start with the free Store Growth Kit, then use the Shopify AI and MCP guide to choose tools around a specific gap. For help reviewing the wider buying journey, book a fit conversation.

Sources and editorial note

• Shopify: Catalog and product discovery
• Shopify: selling on ChatGPT
• Shopify: agentic storefronts
• Google Search Central: product structured data

Sources checked October 20, 2026. The inspiration is a reader-supplied LinkedIn excerpt by Olivia Moore (profile); the original post permalink was not supplied. Her reported shopping results are attributed observations. The audit methods and hypothetical examples here are our editorial recommendations, not measured client results.