AI Platforms

    ChatGPT Shopping: How AI Product Recommendations Work

    How ChatGPT recommends products, where its product data comes from, how checkout in chat changes ecommerce, and what merchants should do to be recommended.

    Saarthak Srivastav

    Saarthak Srivastav

    · 4 min read

    Key takeaways

    • ChatGPT shows product recommendations with images, prices and links when people shop, and OpenAI says these results are organic, not ads.
    • Recommendations draw on structured product data, merchant information and third-party reviews and articles.
    • OpenAI has introduced checkout inside ChatGPT for some merchants and published an open protocol for agentic commerce.
    • Accurate product data, clear product pages and strong reviews are the main levers.
    • Buyers ask in natural language with constraints, so product pages must state attributes explicitly.

    When people shop in ChatGPT, it recommends specific products with images, prices, reviews and links, and in some cases lets them buy without leaving the chat. OpenAI has said these product results are selected organically, not paid placements. For merchants, being recommended depends on three things: accurate, structured product data; product pages that state attributes clearly; and third-party evidence, such as reviews and articles, that the product fits the shopper's need.

    How shopping in ChatGPT works

    A typical shopping conversation:

    1. The shopper describes a need, often with constraints: "waterproof hiking boots under $150 with a wide toe box for day hikes."
    2. ChatGPT searches for matching products, drawing on product data and web content.
    3. It presents options, usually as product cards with images, prices, ratings and merchant links, plus a short explanation of why each fits.
    4. The shopper refines: "which is lightest?", "any vegan options?"
    5. The shopper buys, either by clicking through to a merchant or, where supported, checking out inside ChatGPT.

    In 2025 OpenAI introduced checkout inside ChatGPT for some merchants and published the Agentic Commerce Protocol, an open standard developed with Stripe, for AI-assisted purchases. Merchant programs and requirements continue to change, so check OpenAI's current documentation.

    Where product information comes from

    • Structured product data: feeds and product schema with titles, prices, availability, images, variants and identifiers.
    • Product pages: descriptions, specs, sizing and policies.
    • Reviews and ratings on your site and third-party platforms.
    • Editorial content: buying guides, roundups and comparisons from publishers.
    • Community discussion: forums and threads where people discuss products.

    The pattern matches other recommendation questions: AI looks for consensus across sources. See how AI assistants decide which brands to recommend.

    What merchants should do

    1. Get your product data right

    • Provide complete, accurate product data: title, description, price, currency, availability, brand, GTIN or other identifiers, images and variants.
    • Keep prices and stock in sync with your site. Mismatches erode trust in any shopping system.
    • Follow OpenAI's merchant guidance if you want to share a product feed or support checkout in ChatGPT.

    2. State attributes explicitly on product pages

    Shoppers ask with constraints, so product pages need to answer them in text:

    Attribute type Examples
    Fit and size Wide toe box, true to size, dimensions
    Materials Leather upper, recycled polyester, vegan
    Performance Waterproof rating, battery life, weight
    Use case Day hikes, commuting, small apartments
    Policies Free returns, warranty length, shipping times

    Put these in visible text and in Product schema. An attribute that only appears in an image or a size chart PDF may as well not exist.

    3. Write "who it's for" copy

    ChatGPT explains why each product fits. Give it the language: "Best for day hikes on mixed terrain; not designed for multi-day backpacking with heavy loads." Honest scope makes you a more confident recommendation.

    4. Build review depth

    Ask for reviews steadily, and prompt customers to mention how they use the product and how it performed. Respond to negative reviews. Specific reviews ("comfortable straight out of the box on a 12-mile hike") give ChatGPT concrete details to repeat.

    5. Earn editorial coverage

    Buying guides and roundups from trusted publishers are frequently used for "best [product]" questions. Send reviewers samples and accurate specs.

    6. Make your site readable by AI crawlers

    Allow OAI-SearchBot, serve product content in HTML and keep pages fast. See our AI crawlers guide and how ChatGPT search chooses its sources.

    How checkout in chat changes ecommerce

    When shoppers can buy inside an assistant, some traditional levers weaken: landing page design, on-site upsells and retargeting depend on a visit that may not happen. What matters more:

    • Data quality: the assistant presents your product from your data.
    • Policies: returns, shipping and warranty become part of the comparison.
    • Reputation: reviews and third-party coverage decide the recommendation.
    • Operational reliability: accurate stock and fast fulfillment protect your standing in any marketplace-like system.

    Measuring ChatGPT shopping performance

    • Track product prompts: run shopping questions for your categories and record which products and merchants appear. See AI visibility metrics.
    • Referral traffic from chatgpt.com to product pages, and conversion rates.
    • Orders through any in-chat checkout integration, if you participate.

    A product page audit for AI shopping

    Pick your ten best-selling products and check each one:

    • Title includes the product type and key attribute (for example "Women's Waterproof Leather Hiking Boot, Wide Fit")
    • Price, currency and availability are in text and in Product schema, and match your feed
    • Key attributes (size, material, weight, compatibility) are listed in text
    • A short "best for / not for" statement exists
    • Shipping, returns and warranty are stated on the page
    • Reviews are visible, recent and specific
    • The page loads without JavaScript errors and the content is in the HTML
    • Images have descriptive alt text

    Then ask ChatGPT shopping-style questions that match each product's best use case and see whether it, or a competitor, comes up.

    The bottom line

    ChatGPT shopping recommends products that are well described, well reviewed and easy for its systems to understand. Clean product data, attribute-rich product pages, honest "who it's for" copy, strong reviews and editorial coverage are what get you recommended, and they help on every other AI shopping surface too.

    FAQ

    Frequently asked questions

    How does ChatGPT decide which products to recommend?

    It interprets the shopper's needs, searches for relevant products, and selects options based on how well they match, using product data, merchant information, reviews and other content it finds. OpenAI has said these product results are not ads.

    Can merchants submit products to ChatGPT?

    OpenAI has published guidance for merchants to share product feeds and to integrate checkout through its Agentic Commerce Protocol. Requirements and availability change, so check OpenAI's current merchant documentation.

    What is agentic commerce?

    Agentic commerce is when an AI assistant helps complete a purchase on a shopper's behalf, from finding the product to checking out, often without the shopper visiting the merchant's website.

    Do reviews matter for ChatGPT shopping?

    Yes. Reviews and third-party articles help ChatGPT judge quality and fit, and they supply the descriptive details, such as comfort, durability or ease of use, that it repeats to shoppers.

    • ChatGPT
    • shopping
    • ecommerce
    • product feeds
    • agentic commerce
    Saarthak Srivastav

    Written by

    Saarthak Srivastav

    AEO & GEO, AEO-Rank

    Saarthak Srivastav writes and reviews articles on the AEO-Rank blog, drawing on what the AEO-Rank team sees tracking AI answers across ChatGPT, Gemini, Claude, Perplexity and Grok.