How Online Stores Get Their Products Recommended by AI: How Shoppers Ask, What Product Pages Need, and What Google and OpenAI Officially Say

    By TriloAI Editorial TeamPublished

    When a shopper asks AI for “the best noise-canceling headphones under $100,” ChatGPT and Google's AI features pick a few products from product data, product pages and online reviews, then show prices and links to buy. Online stores need to do three things. Get your product data to the platforms: Google uses Merchant Center, and ChatGPT's product data comes from third-party providers and merchants, with Shopify stores already connected. Write specs, price, who it's for and return terms in text on each product page. And build up reviews from real buyers. Both companies say they don't guarantee your products will be shown.

    Shoppers give AI conditions, rarely brand names

    A common pattern is to put budget, use case and concerns into one question. Here are typical questions online stores run into. They're our examples, not statistics:

    • Recommendations: “best face wash for sensitive skin”
    • Budget: “best noise-canceling headphones under $100” or “dehumidifier under $200 for a studio apartment”
    • Comparing two brands: “Brand A vs. Brand B robot vacuum, which is better”
    • Who it's for: “best phone with large text for seniors”
    • Concerns: “can I return it if I've opened it” or “how long does shipping take”
    • Where to buy: “where to buy authentic [brand] online”

    OpenAI's help center says ChatGPT considers your query and context when it picks products, such as available options, price, reviews and ease of use. If someone mentions a $30 budget, it focuses more on price. If your product page doesn't state a condition like room size in square feet or skin type, AI has one less thing to match. For how the platforms differ, see how AI platforms pick what to recommend.

    ChatGPT's product results aren't ads, and the data comes from product data providers and merchants

    The same help page, updated in October 2026, says product results are selected independently by ChatGPT. They aren't ads and aren't influenced by OpenAI's partnerships. For some eligible products and merchants, shoppers can check out right inside ChatGPT. To decide what to show, ChatGPT considers structured data from first-party and third-party providers (such as price and product description) and other third-party content, answers ChatGPT generated before looking at new search results, and OpenAI's safety standards and product policies.

    When a shopper clicks a product, ChatGPT may list merchants that sell it, ranked by factors like availability, price, quality, and whether the merchant is the maker or primary seller. How the data gets in: for Shopify merchants, product data is already integrated into ChatGPT through Shopify Catalog, with no extra work needed. Other merchants can apply to send OpenAI a direct product feed. Separately, OpenAI's web search page says your site needs to allow its search crawler, OAI-SearchBot, and that placement is not guaranteed.

    Ads are a separate product. OpenAI's shopping page says ads are separate from product results. In a September 16, 2026 announcement, OpenAI said US-based Shopify merchants can use a new ChatGPT Ads app in the Shopify App Store to create and manage ad campaigns, and that the app would be available internationally starting September 23 in markets where ChatGPT Ads are available. We track updates like this in AI search news.

    Google says optimizing for AI search is still SEO, and product data belongs in Merchant Center

    Google Search Central's guide to optimizing for generative AI features (updated July 2026) says that from Google Search's perspective, optimizing for generative AI search is still SEO. Its ecommerce section says generative AI responses can include product listings and product information, and that using Merchant Center (such as Merchant Center feeds) and Google Business Profiles can help your products show up in AI responses and other search results.

    Merchant Center free listings cost nothing. Eligible products can appear on Google Search, Gemini, the Shopping tab, YouTube and more. New accounts are opted in by default, but Google says that doesn't guarantee shoppers will see your products. Required attributes include product ID, title, link, image, price, description and availability, plus brand when there's a clear brand and GTIN when the manufacturer assigned one. In the United States, Google also requires you to set up shipping settings or add shipping costs. Google recommends adding your return policy and keeping prices in Merchant Center matched to your online store. The supported countries list shows the US dollar for the United States and marks the US as a country where the Shopping tab is available.

    The product data specification sets more rules. Titles can be up to 150 characters and descriptions up to 5,000, and both should match your product page. Leave out promotional text like “free shipping,” and keep links, sales information and competitor details out of descriptions. Titles and descriptions written with generative AI go in the structured title and structured description attributes, which let you mark the text as AI-generated.

    Product pages should spell out specs, price, who it's for and return terms in text

    A common pattern is a product page where the specs and size chart live in one long image. Google's advice is to use images and video to support your text. And the price, description, size and material that Merchant Center and structured data ask for are all fields you fill in one by one. If the specs exist only in an image, there's no text to match those fields against. These are the details product pages leave out most often:

    What AI wants to knowWhere to put itHow to write it
    Specs: size, weight, material, capacity, ingredientsProduct page text, Merchant Center, Product structured dataUse a bulleted list or table with units, matching what's on the packaging
    Price and availabilityProduct page, Merchant Center price and availability, Offer price and currencyShow the price in US dollars, the same in all three places. Mark items out of stock when they are
    Who it's for, and who it isn'tProduct page, Product audience (suggested gender and age group), Merchant Center product highlightsGive concrete conditions, like room size in square feet, skin type or age. Don't promise results it can't deliver
    How it differs from other modelsA comparison table on the product pageCompare your own models first. If you compare with other brands, stick to specs you can verify and prove
    RatingsBuyer reviews on the product page, Product aggregateRating and reviewShow only real buyer reviews, and don't filter out the bad ones
    Shipping and returnsProduct page, returns page, Merchant Center return policy and shipping settingsHow long shipping takes and costs, the return window, how to return, who pays return shipping, and which items can't be returned
    Common questionsProduct page FAQ, Merchant Center question-and-answer attributeQuestions customer service actually gets, one answer per question

    Structured data is a standard format in your page code that labels things like “this is a product, this is the price, it's in stock.” Google's product structured data intro describes two types: product snippets for pages where people can't buy directly (such as reviews) and merchant listings for pages where customers can buy from you. For merchant listings, name, image and an Offer are required, and the Offer needs a price greater than zero and a three-letter currency code (USD for US dollars). Google recommends defining your return and shipping policies once under Organization markup. It says providing both structured data and a Merchant Center feed helps Google correctly understand and verify your data, and recommends adding markup to individual product pages, ideally in the initial HTML, because markup generated with JavaScript can make Shopping crawls less frequent and less reliable. Google's generative AI guide also says structured data isn't required for generative AI search. It helps systems identify your products, but it doesn't guarantee AI will cite you. For the technical details, see the GEO technical checklist.

    Shipping times and return terms belong on the page before checkout

    Shipping promises have a federal rule. Under the FTC's Mail, Internet, or Telephone Order Merchandise Rule, when you take orders online you must have a reasonable basis to expect you can ship within the time you clearly state, or within 30 days if you don't state a time. If you can't ship on time, you must offer the buyer the choice to agree to the delay or to cancel the order and get a prompt refund (16 CFR 435.2). The FTC says the rule applies to most goods ordered by mail, phone, fax or online (FTC business guide).

    Returns work differently. The FTC's Cooling-Off Rule, which gives buyers three days to cancel certain sales made at home or at temporary locations, doesn't cover sales made entirely online, by mail or by phone. The FTC also notes that some state laws give consumers more rights (FTC consumer advice). So check your state's rules, then put your terms where shoppers see them before they pay:

    • How long shipping takes and what it costs, matching your Merchant Center shipping settings
    • The return window and the condition items must be in
    • Who pays return shipping, how refunds are issued and how long they take
    • Items you don't accept back, for example custom-made items, perishables or opened personal care products
    • Your business name, contact information and how to raise a problem

    These are also the questions shoppers and AI ask most.

    You're also responsible for what your product pages claim. The FTC says ads must be truthful and not deceptive, and you need a reasonable basis, meaning objective evidence, before you make a claim. Every state also has consumer protection laws that govern ads (FTC Advertising FAQs). Origin claims have a specific rule. Under the FTC's Made in USA Labeling Rule, you can't label a product “Made in USA” unless final assembly or processing happens in the US, all significant processing happens in the US, and all or virtually all ingredients or components are made and sourced in the US. The rule also applies when a mail order catalog or promotional material, including one distributed electronically, carries a Made in USA seal or label. Food, cosmetics, supplements and other regulated products have their own rules, so check with the relevant agency or a professional before you list them.

    Reviews must come from real buyers, and buying reviews or posing as a customer carries legal risk

    OpenAI says ChatGPT's review summaries are based on reviews from public websites, and that OpenAI doesn't verify reviews and ratings. Google's generative AI guide says seeking inauthentic “mentions” isn't as helpful as it might seem, because other systems block spam.

    In the US, the FTC's Consumer Reviews and Testimonials Rule took effect on October 21, 2024. It bans fake reviews, including AI-generated ones; rewards conditioned on a review being positive or negative; reviews by company insiders that don't clearly disclose the connection; and company-controlled review sites that claim to be independent (FTC announcement). Courts can impose civil penalties for knowing violations (FTC Q&A).

    The FTC's Endorsement Guides FAQ adds more. People can't talk about their experience with a product they haven't tried. Connections that shoppers wouldn't expect, such as payment, a free product or employment, should be disclosed clearly and conspicuously. If you send an influencer a free product, ask them to disclose the gift and monitor their posts. Employees who post about your products should say they work for you.

    So don't buy reviews, and don't ask staff, friends or freelance writers to pose as buyers. What you can do is invite every buyer the same way after delivery, not just the happy ones, with no reward for a good review. Some platforms are stricter than the FTC: Google and Yelp ban rewards for reviews outright. For more on US review rules, see how US local businesses get recommended by AI.

    Work through this checklist, starting with your 10 best-selling products

    1. List 10 questions a shopper would ask AI, without your brand name.
    2. Open a Google Merchant Center account, upload your product data, and confirm that free listings show as active. Set up shipping settings, which Google requires in the US.
    3. If you don't use Shopify, follow OpenAI's help center instructions to apply for a direct product feed.
    4. Check that robots.txt and your CDN don't block OAI-SearchBot or Googlebot (method: GEO technical checklist).
    5. In Search Console, open Settings, then the Search generative AI control, and confirm your site is set to be included (that's the default).
    6. On each product page, write specs, price, who it's for and how it differs from your other models, in text.
    7. Add Product and Offer structured data to individual product pages with the currency set to USD, and check for errors with Google's Rich Results Test.
    8. Write one returns and shipping page that covers shipping times and costs, the return window, how to return and which items can't be returned, and make it visible before checkout.
    9. Keep prices and stock consistent across your site, Merchant Center, structured data and any marketplaces you sell on.
    10. Invite every buyer to review the same way after delivery. Don't buy reviews, and don't ask staff, friends or family to write them.
    11. Each month, ask ChatGPT, Gemini and Google AI Mode the same set of questions (method: tracking AI visibility by hand vs. with a tool).

    When you record results, keep “mentioned” (AI names your brand or product) separate from “cited” (AI links to your page as a source). AI answers move around. We asked the same 14 questions twice, and the mention rate came out 31.5% one time and 41.7% the other (see our methodology). Look at the range, not a single run.

    Frequently asked questions

    I only sell on a marketplace like Amazon, Etsy or eBay and don't have my own website. Can AI recommend my products?

    Possibly, but there's no official explanation of how AI treats marketplace listings. OpenAI says that when it lists merchants, it looks at availability, price, quality and whether the merchant is the maker or primary seller. If you have your own brand, your own website gives you the most control over your product data. On a tight budget, start with zero-budget GEO.

    Do I have to pay OpenAI or Google for AI to recommend my products?

    No. OpenAI says ChatGPT's product results aren't ads and aren't influenced by partnerships, and Google's free listings cost nothing. But both say they don't guarantee your products will be shown.

    Can I make opened items final sale?

    That depends on your state's rules and on what you disclose. The FTC's Cooling-Off Rule doesn't cover sales made entirely online, but the FTC notes that some state laws give consumers more rights. Whatever you decide, state it clearly on the product page and in your return policy before checkout, and confirm it with your state attorney general's consumer protection office or a lawyer instead of assuming.

    Does the same approach work if I also sell to Canada or Japan?

    The principles are similar, but AI answers, laws and popular platforms differ by market. Google's merchant listing documentation also says that when you sell in multiple currencies, you should use a distinct URL for each currency, and its own example is a product sold in Canadian and US dollars. For measuring across markets, see measuring AI visibility in other markets.

    Next step: pick your 10 best-selling products, fill in specs, price, who it's for and return terms in text, then ask AI a round of questions without your brand name and write down whether it mentions you and which pages it cites. To find out first whether AI can understand your website, run a free quick scan. It looks at ChatGPT only, needs no sign-up and takes about 1–2 minutes. It checks your website, then runs 2 live test questions in ChatGPT.