A product card is the individual sponsored product unit ChatGPT displays beneath a conversational answer, showing one retailer's item with its image, name, price, stock status, and a Sponsored label. An organic recommendation appears because the model chose it; a product card appears because an advertiser paid and supplied a feed.
If you sell products, the product card is how your catalog becomes clickable inside a ChatGPT answer, so you must decide whether to give OpenAI the feed that lets your items appear. Skip that feed and a rival's card fills the shopping moment while your brand stays invisible in the conversation where the buyer is choosing.
Product cards are the ad creative ChatGPT assembles from your product feed the moment a conversation turns commercial. Each card represents one item and renders below the answer, clearly labeled as sponsored and visually separated from the organic text. You do not design the card by hand; you supply catalog data and OpenAI's system decides when to build a card and which item to show.
A single card draws on the structured fields in your feed: title, description, image, price, availability, and identifiers such as GTINs. OpenAI reuses the Google Shopping schema, so the same file you send Google Merchant Center can populate cards here. For a card to appear, the matching product has to be eligible, the feed has to be live, and the conversation has to signal buying intent.
Product cards sit inside the ChatGPT Ads system, one step below the campaign and feed that govern them and one step above the carousel that groups several cards together. AirOps tracks which product cards and advertisers surface in the ChatGPT answers you already monitor, so you can see your cards next to competitors' in the same shopping moments.
Resources: see how AirOps connects ChatGPT ad activity to your pipeline data
A product card exists only when several things line up in sequence, from the feed you upload to the answer a shopper sees.
Upload the feed. You send a product catalog to ChatGPT Ads Manager using the Google Shopping file format, with titles, images, prices, and availability for each SKU.
Match intent. When a user's conversation signals commercial intent, OpenAI's system scans eligible products and scores them against the context of that prompt.
Assemble the card. The system builds a card from your feed fields, pairing the product image with its name, price, and stock status, and attaches the Sponsored label.
Place the unit. The card renders below the answer, either alone or grouped with other cards as a carousel, a choice OpenAI's platform makes.
Route the click. A tap sends the shopper to your product page or, for eligible catalogs, toward an in-chat checkout.
The placement tells you which products OpenAI judged relevant to a buying conversation and whether your feed was eligible. It does not tell you which prompt triggered the card or which rivals appeared beside you.
Resources: work through targeting, creative, and ad-policy best practices for ChatGPT ads
Product cards decide whether your catalog is even in the room when a shopper asks ChatGPT what to buy. That makes feeding the channel a real budget decision, because the paid shelf is filling up fast. AirOps research found ChatGPT ads went from appearing in 0% of answers to more than 50% within three months, a 113x increase in AI ad load.
They capture high-intent moments. A product card lands exactly when someone is comparing options, and Circana's March 2026 Shopper Survey found 25% of consumers had used an AI recommendation to buy a consumer packaged goods product.
They protect share of the shelf. If your feed is missing or ineligible, a competitor's card fills the placement and you lose the sale without ever seeing the prompt that drove it.
They tie paid to organic. The same conversations that trigger a card are the ones your organic answers compete in, so product cards let you cover a query with both paid and earned presence.
SEO managers use product cards to defend branded shopping queries in ChatGPT that their organic pages can no longer reach on free and Go plans.
Growth marketers use product cards to put bestsellers in front of high-intent shoppers at the moment a conversation turns to buying.
Demand gen leads use product cards to extend catalog coverage across the AI conversations where competitors are already bidding for attention.
A product card can only be built from a structured catalog file, so the completeness and accuracy of your feed's titles, images, prices, and availability directly sets what a card can say about that item on screen.
Each product must pass OpenAI's policy and data checks to be shown, and an item that is out of stock, mispriced, or missing required fields quietly drops out of the running before a shopper ever sees it.
Unlike keyword ads, cards surface when the model reads a conversation as commercial, so the match depends on the context and wording of the exchange instead of a fixed query you bid on in advance.
Appear in ChatGPT shopping answers that paid plans and organic pages cannot reach.
Reuse your existing Google Shopping feed, so setup leans on a file you already maintain.
Capture higher-value demand: Adobe Analytics found AI-referred shoppers converted at a rate 60% higher than non-AI traffic on US retail sites in July 2026.
Show live price and stock, so the card reflects what a buyer can actually purchase.
Let OpenAI's system match your items to intent, widening reach beyond keywords you pick.
Keep your feed clean and current, because a card inherits every error in your product data.
Write product titles that read well on their own, since the title carries most of the card's meaning in a small space.
Use clear, single-product images, so the card is legible the instant it renders below an answer.
Keep price and availability synced in real time, so cards never show items a shopper cannot buy.
Send clicks to a matching product page that OpenAI's crawler can reach, so the destination aligns with the card.
Start with your top SKUs by revenue, so the feed sample represents your strongest catalog first.
Avoid treating the card as set-and-forget. Competent teams upload a feed, see cards appear, and then let stale prices and stockouts pile up until the card advertises products that are gone, which burns spend and erodes trust in the exact moment a buyer is deciding.
AirOps: tracks which product cards and advertisers appear in your monitored ChatGPT answers and connects that paid activity to your organic AI search data.
ChatGPT Ads Manager: OpenAI's native platform at ads.openai.com where you upload a product feed, set bidding, and launch the campaigns that generate cards.
Google Merchant Center: manages the structured product feed whose Google Shopping format OpenAI reuses, so the same catalog can power cards in ChatGPT.
Audit your feed. This week, open your existing Google Shopping feed and check that titles, images, prices, and availability are accurate for your top products, with no approval needed before you spend anything.
Create an account. Sign up for ChatGPT Ads Manager at ads.openai.com and confirm your business is eligible to advertise in your market before building anything larger.
Upload a sample. Submit a sample of your catalog, starting with your highest-revenue SKUs, so OpenAI can begin generating cards from real products. A hundred of your best products is a strong starting sample.
Set bidding and destinations. Choose your campaign objective and bidding, and point each card's clicks to a matching product page the crawler can reach.
Monitor and refine. Watch which cards appear and against which competitors, then tighten feed data and bids based on what actually surfaces in real ChatGPT answers over the first weeks.
A product card is a sponsored, single-product unit ChatGPT builds from your feed and shows below an answer.
Cards are generated automatically from catalog data, so you influence them through the feed instead of by designing each ad.
Your feed is the main constraint: no eligible, accurate catalog data means no card appears.
The main risk is stale data, where a card promotes a product that is out of stock or mispriced.
The leverage sits in feeding and cleaning the SKUs tied to your highest-intent shopping queries inside ChatGPT.
A product card is paid; an organic recommendation is not. When ChatGPT names a brand or product inside its own answer text, that mention is editorial, chosen by the model from what it has learned and retrieved. A product card is a distinct, labeled ad placement that sits below the answer, built from a catalog an advertiser submitted and governed by bidding and budget. The practical difference is control and eligibility. You cannot pay your way into an organic mention, and you cannot earn a product card without a live, eligible feed. They can also appear together: ChatGPT might describe a category in its answer and then show sponsored cards underneath, so a shopper sees both an unpaid opinion and a paid option on the same screen. Treating them as one thing leads to bad decisions, because you improve organic mentions through content and authority and you improve cards through feed quality and bids.
A product card appears when OpenAI's system reads the conversation as commercial and finds an eligible product in your feed that matches the moment. There is no fixed schedule or guaranteed frequency; placement is decided in real time, answer by answer. Ads currently serve to logged-in and, in some cases, logged-out users on the free and Go plans, so paying subscribers on higher tiers do not see them at all. That means the same prompt can trigger a card for one user and nothing for another, depending on plan, region, and context. You also do not choose between a single card and a carousel, because OpenAI decides whether to show one product or several. Practically, a card is most likely when a shopper asks for a product, a comparison, or a recommendation in a category you sell into, and when your feed has a relevant, in-stock, policy-compliant item ready to serve.
Product cards vary because the system scores your products against the exact wording and context of each conversation, and that context shifts constantly. Two prompts that look similar to you can read very differently to the model, so one surfaces your card and the other does not. AI product responses are also unstable by nature: an audit published on arXiv in 2026 found ChatGPT expressed a first-person product preference in 79% of its product-recommending responses, and the specific products it named often changed across repeated requests. Eligibility adds more variance, since a product can drop out the moment it goes out of stock, breaches a policy, or loses a required feed field. Bidding and competition move the result too, because other advertisers are scored in the same moment. The honest takeaway is that you should expect variance and measure cards across many runs instead of reading a single answer as the whole picture.
You can influence product cards, but you cannot fully control them. You control the inputs: which products you submit, how accurate and complete the feed is, your bids and budget, and the landing pages clicks go to. You do not control the output, because OpenAI's system decides when a card is eligible, which item to show, whether to group it into a carousel, and where it ranks against competitors. So the honest answer is no, you cannot guarantee a specific card on a specific prompt. What you can do is raise your odds: keep every high-priority SKU in stock and correctly described, make titles and images carry the message, and bid competitively on the categories that matter. Your real job is to stack the deck through feed quality and bidding, then let the system match your items to the conversations where they fit best.
A good result depends on your goal, but the clearest benchmark is incremental and profitable sales instead of raw impressions. Because cards appear in high-intent shopping moments, judge them on conversion rate, cost per acquisition, and return on ad spend, the same way you would judge Google Shopping. A healthy card program shows your products surfacing consistently for the queries you care about, clicks landing on fast, matching product pages, and a cost per sale your margins can sustain. Share of the shelf matters too: if competitors' cards appear for your core queries and yours rarely do, that gap is a problem even when your absolute numbers look fine. Early on, prioritize coverage and clean data over efficiency, because you need enough card impressions to read performance. Once volume is steady, tighten bids and feed quality toward the SKUs and queries that return the best margin.