A shopper asks Google's AI Mode for a waterproof dog jacket under $60 that ships before Dec. 20, and a small online store selling exactly that jacket never appears in the answer. The jacket and its price are fine, but the store's product page describes the shipping window in a paragraph of prose and puts the return policy behind a footer link. Software reading that page has no structured way to confirm details a human shopper could find in five seconds, so it moves on to listings that state them plainly.
Adobe forecasts that traffic from AI services to U.S. retail sites, counted by shoppers clicking a link, will rise 130 percent year over year from Nov. 1 through Dec. 31. Google, meanwhile, spent this year building checkout directly into its AI shopping surfaces, so an agent can now carry some shoppers from a question to a completed order. A small store that wants a shopping agent to recommend its products needs clean structured data, a complete product feed, and shipping and return terms that software can parse without guessing.
The fixes are modest and mostly free, and a store that starts on Thursday, Nov. 19, has eight days to finish them before Black Friday on Nov. 27.
On Jan. 11, at the National Retail Federation's annual show, Google's announcement of the Universal Commerce Protocol called it "a new open standard for agentic commerce that works across the entire shopping journey." Shopify, Etsy, Wayfair, Target and Walmart co-developed the protocol, and more than 20 other companies endorsed it, including Stripe, Visa, Mastercard, American Express and Best Buy. Speaking that day as what he called an indecisive shopper, Google chief executive Sundar Pichai said he looks forward to "the day when agents can help me get from discovery to purchase."
A Sept. 16 Google post on holiday shopping updates said UCP "already enables direct checkout for hundreds of thousands of brands and retailers across Google." The same post said cart transfer to merchant sites is rolling out gradually in the United States, with Australia and Canada scheduled to follow early next year. It also reported that merchants who follow core Merchant Center feed best practices see conversions rise an average of 5 percent in the month after they adopt them.
OpenAI has approached the same problem from a different direction, according to an April analysis from the payments firm Checkout.com, which found that the company moved away from completing purchases inside ChatGPT. In the models OpenAI now favors, the assistant handles discovery while checkout runs on the merchant's own site or app. For a small seller, the two approaches require the same preparation, because both depend on an agent understanding a product correctly before any shopper reaches a cart.
An AI shopping assistant reads a page's structured data, the product feed a store submits to platforms like Google Merchant Center, and the visible text, then checks all three against the shopper's constraints. Price, stock status, size, delivery date and return terms correspond to a property in schema.org markup, and Google's merchant listing documentation names the ones it uses, from offers and availability to shippingDetails and hasMerchantReturnPolicy.
Most small stores already emit some Product markup, because WooCommerce, Shopify and the major SEO plugins generate a basic version automatically. The fields those automatic versions leave empty are usually the ones that answer holiday questions. A typical page declares a price and an in-stock flag while saying nothing machine-readable about shipping cost, handling time, delivery speed or how many days a buyer has to return a gift. Those are the details an agent has to confirm before it can tell a shopper that a jacket will arrive by Dec. 20.
A store owner can run the audit in an afternoon by putting three best-selling product URLs through Google's Rich Results Test and noting which merchant listing fields come back missing or flagged. The next step is to compare what each page says in plain text with what its markup claims, line by line. A sale price in the copy paired with the old price in the markup is the kind of inconsistency that can lead an automated system to skip a listing entirely.
Google Merchant Center accepts free product listings, and a store that has never submitted a feed does not appear in the parts of Google's shopping surfaces that draw from one. Shopify and WooCommerce both offer official or widely used integrations that sync a catalog automatically, which keeps prices and stock levels current without manual uploads during the busiest weeks of the year.
Google's September post also pointed merchants toward newer feed fields, including conversational attributes that add context to listings and a video link attribute for shoppable formats. A small store does not need all of them before Black Friday, because a short list of basics covers most of what an agent checks first. Those basics are accurate titles, complete descriptions written for a person rather than a keyword list, correct GTINs where products have them, and shipping and return settings filled in at the account level.
Shoppers now describe what they want in full sentences, and agents compare those sentences against product copy to decide which listings fit a request. A description reading "Premium jacket, multiple sizes, great for winter" gives an assistant almost nothing to work with. A description that names the waterproof rating, the size range, the weight, the materials and the breeds it fits hands the same assistant a dozen specific facts to match against a shopper's words.
Small sellers hold an advantage over marketplaces at this step, since the owner of a small shop knows details that no catalog import supplies. That owner knows the medium runs small and gets returned, which colors sell out first, and which item people buy as a gift for a new puppy owner. Written into the description in everyday language, that knowledge becomes material an agent can cite when it answers a shopper. Our earlier guide on AI crawler controls for small sites covers how to let the assistants you want reach your pages in the first place.
When an AI system tells a shopper that your jacket arrives by Dec. 20 and costs $54 with free returns, it is repeating what your data said. If that data is stale, the shopper blames your store for the late package or the surprise return fee, not the assistant that relayed the claim. The same principle applies to any automated tool that speaks on behalf of a business. Our piece on website chatbot liability examined a tribunal ruling that held a company to what its own chatbot told a customer.
Set a holiday cutoff date for standard shipping and publish it in the feed, the markup and the visible page at the same time, so every source gives an agent the same answer. When stock runs low, update the stock status in the feed and the markup instead of leaving a sold-out item listed as ready to ship. Adobe expects Cyber Monday on Nov. 30 to bring in about $15.1 billion in U.S. online sales, and every order placed that day will test the claims a store's data made the week before.
The first four of those eight days can go to testing top products and filling missing shipping and return fields in the markup and in Merchant Center. That stretch should end with a check that every price matches across the page, the markup and the feed. The remaining four days can go to rewriting descriptions for the ten products most likely to be bought as gifts, so each one answers the questions a real shopper asks out loud.
None of this requires a developer or a new subscription, only an owner willing to read each product page field by field, the way a machine does, and fix what is missing. AI shopping traffic is forecast to more than double this year, and it will reach the stores whose product structured data matched their real prices, stock and shipping dates on Nov. 27.