Google AI Mode Shopping: Merchant Center Fixes for a Small Store

A small store should fix six things in Google Merchant Center before expecting its products to show in Google AI Mode: a verified account with free listings on, complete required product data, correct GTINs or brand and MPN, accurate images, shipping settings, and a return policy that matches the website. Google does not guarantee placement.

Google describes AI Mode shopping as answering a shopper's conversational question with prices, reviews and stock details from its Shopping Graph. That graph is built from product listings, and Google's own AI optimisation guide says Merchant Center feeds can help products be visible in AI responses, though it does not make a feed a condition. An assistant can audit your data for each fix, as long as it is never allowed to invent a value.

Follow me on Instagram@sagnikteaches

What is verified and what is not

Most pages about AI Mode shopping come from agencies and feed tools, and they state requirements more firmly than Google does. It helps to separate what Google's own pages say from what others claim.

Connect on LinkedInSagnik Bhattacharya

• Google says that its shopping experience in AI Mode returns organised answers with prices, reviews and inventory details, and that the Shopping Graph covers more than 50 billion product listings, with around 2 billion refreshed every hour.
• Google's free listings help page says products need required attributes (ID, title, link, image link, price, description and availability), shipping information, and a return policy that it recommends adding in Merchant Center. It adds that Google does not guarantee products will show, because matching depends on how well your data fits what customers search for. That page lists the Shopping tab, Search, Images, Lens and YouTube as surfaces and mentions Gemini in its introduction; it does not name AI Mode.
• Google's guidance on AI features says there are no extra requirements or special optimisations to appear in AI Overviews or AI Mode: no special schema.org markup and no AI text files such as llms.txt, which Google Search ignores. Its AI optimisation guide adds that a site must be included in generative AI features in Search Console.
• Others claim that a feed is mandatory for AI Mode or that structured data on the product page is a requirement. Neither appears on Google's pages as a condition, so treat such claims as unverified.

Checkout inside AI Mode is a separate, optional programme. Google's Merchant Center help describes a Universal Commerce Protocol (UCP) integration hub that adds a checkout button to eligible listings in AI Mode and on Gemini: a merchant completes a technical integration, fills in an interest form and waits to be selected, and the rollout is gradual and limited to a few markets. Merchants on a platform are told to ask that platform about its plans. Nothing on that page makes checkout a condition for appearing in AI Mode, so do not plan around it. Plan around being accurate, complete and eligible, which helps on every Google surface and costs nothing but time.

Subscribe on YouTube@sagnikteaches

Fix 1: an account that Google will trust

Before the product data, check the account itself. Disapprovals at this level take every product with them.

1. Verified and claimed website. The domain in your product links must be the one verified in Merchant Center. Google's specification asks for landing pages on your verified domain.
2. Free listings switched on. Google says the feature is on by default in most cases; check the Free listings card under Marketing methods, and note that changing it needs Admin access.
3. Business information that matches the site. Name, address and contact details that appear on your pages should agree with the account. A mismatch is a common starting point for the misrepresentation problems covered in fixing a suspended Merchant Center account.
4. Policies visible without a login. Returns, shipping, privacy and contact pages should be reachable from the footer and readable by a visitor who has not signed in.

Fix 2: the required attributes, filled in properly

Google's product data specification requires, for every product, an id, a title, a description, a link, an image link, an availability value and a price. It asks for brand on new products (except movies, books and music), and applies extra attributes for apparel such as colour, size, age group and gender. The limits are plain: a title up to 150 characters, a description up to 5,000, and an id of at most 50 characters that stays the same across updates.

The quality traps are specific:

• Title. It should describe the product accurately, match the landing page and avoid promotional text or all capitals. "Walnut desk organiser, 3 compartments, 28 cm" is better than "BEST SALE!!! Desk Organiser FREE SHIPPING".
• Description. Describe only the product: no store links, sale text or competitor names.
• Price. The submitted price and currency must match the landing page and checkout. For a sale, the page and the feed need to agree on the same day.
• Availability. One of in stock, out of stock, preorder or backorder, matching the page. A preorder needs an availability date within a year.
• Brand. The brand shoppers recognise. Use your own only if you make the product or sell it under your label, and never enter "Generic" or "N/A" as a placeholder.

Let AI audit the export, not rewrite it blind

Export your product feed or catalogue to a spreadsheet and give the assistant the columns. Ask for a report, not a rewrite:

Here is an export of my product feed (columns: id, title, description, link, image_link, availability, price, brand, gtin, mpn, condition). For each row, list problems against these rules only: title over 150 characters; title with promotional text or all-caps words; description over 5,000 characters or containing a URL, sale wording or shipping wording; missing brand; brand equal to "Generic" or "N/A"; missing gtin AND missing mpn; price of 0 or without a currency code; image_link that does not start with http or https. Return a table: id, rule broken, the exact text that breaks it. Do not fix anything and do not invent any value.

An illustrative result for a 40-product store might list 9 titles with promotional text, 6 missing brands, 12 products with neither a GTIN nor an MPN and 3 descriptions with a shipping line. That is a short, ordered to-do list. Check a sample of the flagged rows by hand, because an assistant can misread a column or flag a title that is fine, and a wrong flag wastes the time you were trying to save.

Fix 3: identifiers, where mistakes are most expensive

Identifiers are the item on this list where AI is most likely to cause harm if misused. Google's specification states that a GTIN should be submitted only if you are sure it is right, because an incorrect one gets the product disapproved, and that you must not guess or generate one.

The rules, as Google's specification puts them:

• A GTIN is up to 14 digits (Google ignores dashes and spaces), must have a valid check digit, and must not use restricted prefixes such as 02, 04 or 2, or coupon prefixes 98 and 99.
• For products made by someone else, use the GTIN and brand of the manufacturer that actually made it.
• An MPN is needed only if the product has no manufacturer-assigned GTIN, and must be the manufacturer's own value, specific to the variant (different colours need different MPNs).
• identifier_exists may be set to no only in the cases Google lists, such as a product with no GTIN and no valid brand and MPN combination. Do not set it to no for a product that has identifiers.
• Products submitted without a GTIN where one exists may have limited visibility.

Where do GTINs come from? From the manufacturer's packaging, the supplier's spreadsheet or the barcode on your own products if you registered them. An assistant can help you find which rows lack one and draft the request to your supplier, but it must not supply the digits. The tutorial on what AI can and can't fill when a GTIN is missing goes through the cases in detail, including handmade and own-brand goods.

Fix 4: images that pass and sell

Google's specification asks for images of at least 500 by 500 pixels, with enforcement of that minimum beginning on 31 January 2027. Images should accurately show the product and carry no promotional text, watermarks, borders or placeholder graphics, and must not be scaled-up thumbnails. Google must be able to crawl the image URL, so robots.txt must allow its crawlers.

Check three things on every product: the main image shows the product itself on a plain or neutral background, there are extra angles in additional image links (up to 10), and no image is a "coming soon" placeholder. If you create or alter images with AI, Google's specification asks you to keep embedded metadata such as the IPTC digital source type tag; the fuller picture is in Merchant Center image requirements for AI-generated product photos.

Fix 5: shipping data that tells the truth

Google's free listings page says to set up shipping, either in Merchant Center's shipping settings or with the shipping attribute on each product, and notes that it is required in a long list of markets. A shopper in AI Mode is comparing, so unclear or missing delivery costs lose you to a store that states them.

• Set rates by service and region the way your store really charges them, including free-shipping thresholds.
• Give a realistic handling time and transit time. A promise your couriers cannot keep leads to complaints that count against you.
• Make sure the cost shown at checkout matches the cost in Merchant Center. A mismatch between feed and checkout is a recognised reason for disapproval.

If you make to order or ship from several places, write the settings out in a table first (service, region, cost, handling time, transit time) and check it against your own shipping page before entering it.

Fix 6: a return policy in both places

Google describes returns as optional but recommends them, and a return policy is part of the free listings guidance. In Merchant Center it is set under Products & store, then Shipping and returns, then Return policies and Add return policy. The flow asks for a policy URL, countries, whether you accept returns (for defective items only, or all), exchanges, product condition, return window in days, the return method, how the return label is provided, any restocking fee and the refund processing time. Google reviews each policy and shows a status of Verified, Pending or Rejected. Without an uploaded policy, Google may show a general one it collected from your website, labelled "For most items", so it pays to state yours.

The important rule is consistency: the clear returns and refund information on your website, readable without signing in, has to match what you enter in Merchant Center, including the footer, returns page and any banners. A thirty-day window in Merchant Center and a fourteen-day window on the website is a contradiction Google will treat as one. If some products differ, such as made-to-measure items, Google lets you add an exception policy with its own label and attach that label to those products.

AI as a consistency checker

Below are (A) my return policy as entered in Merchant Center and (B) the text of my website's returns page. List every difference in: return window, who pays for return shipping, restocking fees, which products are excluded, and refund timing. Quote the exact words from each source side by side. Do not decide which is correct.

Ask for exact quotes so you can verify each difference yourself.

The Shopify route, step by step

If your store runs on Shopify, the usual path is the Google and YouTube channel. Shopify's guidance for AI shopping on Google follows this order:

1. Install the channel, connect your Google account and sync your product feed. The sync keeps price, availability and product data updated in Merchant Center.
2. Complete the product fields and identifiers. Shopify advises filling every required and recommended field for each listing, and notes that Google also requires certain classification fields to categorise products.
3. Write clear listings: titles with brand, type and key differences; descriptions that say when and how the product is used; high-resolution images on plain backgrounds from several angles.
4. Collect and respond to reviews. Shopify says a review app can make ratings eligible for Google's shopping results.
5. Check the Diagnostics tab in Merchant Center, fix problems in Shopify admin, and sync again.

Two cautions. Shopify's advice is to correct product data in Shopify admin and sync again, so the store and the feed never disagree. And Shopify describes its agentic storefronts as early access with varying availability, so read what Shopify's agentic storefronts do and what you still control before assuming any setting applies to your store.

For titles, the tutorial on rewriting Google Shopping titles with AI safely sets out rules that keep a rewrite from drifting away from the product.

Worked example: a 40-product store, one afternoon of fixes

This is an illustration with invented figures. A small store sells homewares: 40 products, 120 variants, on Shopify with the Google channel installed two years ago and rarely checked. Diagnostics shows 31 of 120 variants with warnings and 9 disapproved.

1. Export (10 minutes). The owner exports the product list and the Merchant Center issues report.
2. Audit (15 minutes). The assistant runs the rules from the audit prompt over the export. It reports 9 promotional titles, 14 variants with no GTIN or MPN, 6 products with a placeholder image, 3 with a price of "0" on a bundle, and a description with a link to a sale page.
3. Triage (20 minutes). The owner checks 10 flagged rows by hand and agrees 9. One title the assistant flagged is acceptable. Of the 14 identifier gaps, 8 are products with manufacturer barcodes on the box and 6 are the store's own handmade line.
4. Fixes (about 2 hours). Barcodes are typed from the physical packaging, not from memory. For the handmade line the store is the manufacturer, so its own brand and product code go in as brand and MPN; those rows have identifiers and are not marked identifier_exists no. Placeholder images are replaced by real photos. Titles are rewritten from the product's own facts.
5. Shipping and returns (30 minutes). The shipping table is rechecked against checkout, and one free-shipping threshold that differed by 5 is corrected. The returns window in Merchant Center is 30 days, the website says 28: the website is changed to 30, since the policy actually honoured is 30.
6. Sync and wait. After the next sync, warnings fall from 31 to 4 and the disapproved count to 1, over the next day or two.

Nothing in that list guarantees a single appearance in AI Mode. It removes the reasons a product would be excluded from Google's shopping surfaces, which is the part within the owner's control.

A price that was right on one page and wrong on another

An illustrative failure that Shopify's own advice flags as a common cause of disapproval. A store ran a weekend sale by changing a compare-at price and a sale price in its theme, but a currency-conversion setting rounded the feed's price to a different figure from the product page. For two days 20 products carried a price in the feed that was a few units higher than the one shown at checkout. Google's specification asks for the price to match the landing page and checkout, and the mismatched products were disapproved.

The cause was invisible in the store, because every page looked right. It was found only in Diagnostics, where the issue named the price mismatch. The routine that prevents it: after any sale or price change, wait for the sync, then open three product pages and compare them with the Merchant Center product list.

A weekly routine that takes twenty minutes

1. Open Diagnostics and note new warnings and disapprovals. Fix the largest group first, since one cause often explains many products.
2. After any price, stock or theme change, compare three live pages with the feed values.
3. Run the audit prompt on a fresh export once a month, and the policy-consistency prompt whenever you edit returns or shipping.
4. Ask for reviews after delivery in a way that follows Google's product ratings policies: an incentive must not depend on what the review says and must be disclosed in the reviews feed, and every review, including low-star ones, goes to Google.
5. Keep a log of changes and dates, so a drop in visibility can be matched to what you changed.

The page-side basics matter too. Google says AI features need no special markup, but accurate structured product data and a clear page help every search system. If you are working on descriptions in bulk, generating product descriptions from CSV rows covers keeping them factual.

What to expect, honestly

Being eligible is not the same as being shown. Google decides what to display for each shopper's question based on its own matching, and you cannot buy or force a place in AI Mode answers. What you can do is make sure every product has complete, accurate, current data, so that when a query matches, your product is a candidate and not a disapproval.

It is also worth measuring what you can. Note impressions and clicks from free listings in Merchant Center before and after your fixes, over several weeks, and expect small numbers if the store is small. If you want help working through an export of your own catalogue, that is the kind of job my AI implementation consultation covers. Re-read Google's product data specification and free listings page once a quarter, since specifications and dates change, as the image rule beginning in January 2027 shows.

Questions small stores ask about AI Mode and Merchant Center

Do I need Merchant Center to appear in Google AI Mode shopping?

Google's pages do not make a Merchant Center feed a condition for AI Mode, and no one can promise placement. But Google's AI optimisation guide says Merchant Center feeds can help products be visible in AI responses, and its free listings help page says showing depends on how well your data matches searches. A clean feed is the sensible starting point.

Do I have to run Shopping ads?

No. Google's free listings help page describes products appearing at no cost, and says the Free listings feature is on by default in most accounts. Ads are a separate choice. You do need an approved feed and the shipping and returns information Google asks for.

What if some of my products have no GTIN?

Do not invent one. Google's data specification says an incorrect GTIN gets a product disapproved. If the product has a manufacturer GTIN, submit it exactly. If it has none, provide brand and MPN where they exist, and set identifier_exists to no only where Google's rules say you may.

How often should I check Merchant Center?

Look at the Diagnostics area at least weekly, and after every price change, bulk upload or theme change. Price mismatches, missing identifiers and placeholder images are among the usual causes of disapproval, and a quick weekly look keeps a small problem from taking half the catalogue offline.

Further reads

• Perplexity Merchant Program: Should a Small Online Store Join? — Another AI shopping route and whether it suits a small store.
• ChatGPT Shopping Agent: Can It Find a Small Online Shop's Products? — Can ChatGPT find a small shop's products?
• Judge.me vs Yotpo: Are AI Review Summaries Worth Paying For? — Reviews feed product ratings; compare two review apps.
• How to Get Your Products Recommended in ChatGPT Shopping — The same job for ChatGPT's shopping results.
• Performance Max vs Standard Shopping for a Small Online Store — If you later add paid Shopping, pick the campaign type.
• AI Online Shopping Solutions for a Small Store: Chat, Quiz or Search? — Chat, quiz and search tools on your own store.
• Help ChatGPT Discover Your Products: A Checklist for Small Shops — A grouped checklist for small online shops, with a scored audit of twenty products, test prompts and the feed fields OpenAI documents.
• Google Ads for Dropshipping With AI: What to Let It Write — What AI can safely draft for a dropshipping store's Google Ads, what only you can verify, and the budget sum that decides whether to start.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: Google Merchant Center Help (product data specification, free listings, return policies, product ratings policies, UCP integration hub), Google Search Central (AI features, AI optimisation guide) and Google's shopping announcement on its news pages, read November 2026; Shopify's published guidance on Google AI shopping. Check each page again before acting, as Google changes specifications and rollout rules often.