ChatGPT Product Feed Ads: What Retailers Need to Know in 2026

On May 12, 2026, OpenAI shipped product feed ads inside ChatGPT. Retailers can now connect a product catalog to the platform, set filters for which items are eligible, and let ChatGPT auto-generate sponsored placements from product names, images, and catalog attributes. The placement itself looks unchanged to the user: a sponsored card below the AI's answer, clearly labelled. What changed is everything behind the placement.

For ecommerce brands, the May 12 announcement is the inflection point. Running ChatGPT Ads at catalog scale was impractical until now because campaigns had to be built one product at a time. The platform can now handle up to 1 million SKUs per advertiser. A mid-market retailer with 50,000 items can actually participate. The piece below covers what shipped, what the new eligibility test looks like, and what catalog work retailers should start immediately.

What ChatGPT Product Feed Ads Are

Product feed ads are sponsored placements auto-generated from structured catalog data, not built one at a time. You connect your product feed, and ChatGPT's system creates ad cards from the product information in that feed: title, image, price, availability, and descriptive attributes.

The mechanics are similar to Google Shopping and Meta Advantage+ Shopping. You maintain a single product feed, and the platform decides which products to surface based on the shopper's query. Criteo is the launch ad-tech partner handling feed ingestion and serving. StackAdapt has also confirmed feed compatibility for advertisers already running through that platform.

The ads appear as sponsored product cards within the conversation, clearly labeled as sponsored content. They sit below ChatGPT's conversational response, so the shopper sees the AI's answer first and the sponsored products second. This placement means the ad needs to be relevant to what the shopper actually asked, not just broadly related to a keyword.

The 100-product Sample Is The New Eligibility Test

OpenAI requires new partners to submit a roughly 100-product sample before granting full catalog access. The sample must pass review for data completeness, image quality, and attribute accuracy.

This is a meaningful departure from how Google, Meta, and Amazon handle feed onboarding. On those platforms, you connect the full feed immediately and fix issues as they come up. OpenAI checks quality before you spend, not after. That means catalog quality is now a prerequisite for paid placement, not an optimization you get around to later.

The review process filters for specific problems. Products with missing attributes, low-resolution images, or inaccurate pricing data will not pass. Generic titles like "Product 12345" or "Men's Shoe" without further detail will also flag issues. The bar is not unreasonably high, but it does require that your feed be actively maintained rather than exported once and forgotten.

For retailers who have been running Google Shopping with a "good enough" feed, the 100-product sample is a forcing function. It pushes you to fix the data quality issues you have been carrying for years. That cleanup work benefits every channel your feed touches, not just ChatGPT.

Your Google Shopping Feed Works On Both Surfaces

One structured feed serves Google Shopping, ChatGPT paid placements, and ChatGPT organic recommendations. No reformatting is required to move your existing Google Merchant Center feed into ChatGPT's ad system.

That said, "compatible" and "effective" are different things. A feed that meets Google Shopping's minimum requirements can still underperform on ChatGPT. The reason is how matching works. Google Shopping matches on keywords: title, price, and a handful of core attributes do most of the work. ChatGPT matches on conversational intent, which means descriptive attributes carry more weight.

A shopper on Google types "gold bracelet under $200." A shopper on ChatGPT asks "what is a good anniversary gift under $200 in gold?" The second query needs your feed to include occasion, material, and gift-related attributes. If those fields are empty, the product will not match.

The upside is that feed investment pays off across channels. Every attribute you add, every title you rewrite to be more descriptive, and every image you upgrade improves performance on Google Shopping, ChatGPT ads, and organic AI agent surfaces simultaneously. The work is not duplicated across platforms. It is shared.

For a broader look at how ​search engine optimization and AI visibility intersect for ecommerce brands, Passionfruit's SEO service covers both traditional feed optimization and the AI readiness layer.

The Gap Between Seeing AI Agent Traffic And Being Ready For It

95% of merchants already see AI agent traffic on their sites. Only 20% have machine-readable catalogs those agents can actually use. That 75-point gap is where most retailers sit: agents are showing up, but the catalog data is not structured well enough for the agent to reliably recommend the right product. Fixing this is largely the same work as preparing a product feed for ChatGPT's paid ads: clean titles, accurate attributes, current pricing and availability, and valid schema markup.

Passionfruit's ​AI Visibility Tracking measures whether your AI agent catalog readiness is actually being surfaced, not just whether your feed is technically valid.

What Good Catalog Data Means For Conversational Matching

Conversational matching relies on descriptive attributes that most product feeds treat as optional. Material, occasion, finish, intended recipient, use case: these are the fields that connect a natural-language question to the right product.

Many catalogs fill in the required fields (title, price, image, availability) and leave everything else empty. That worked when the primary discovery channel was keyword search. It does not work when shoppers ask questions like "what is a good anniversary gift under $200 in gold," and the AI needs to match on occasion (anniversary), material (gold), and price (under $200) simultaneously.

Consider a jewelry retailer with 2,000 SKUs. If the feed has title, price, and image but leaves the "occasion" and "material" fields empty, the retailer's products will not match conversational queries that include those attributes. The products exist. The data to match them does not.

Before submitting your 100-product sample, audit attribute completeness across your full catalog. The audit should answer two questions: which attributes are empty, and which attributes are filled but inaccurate? A material field that says "metal" when the product is 14k gold is nearly as unhelpful as an empty field.

The retailers who get the most from ChatGPT product feed ads will be the ones who treat their catalog data as a product in itself, something that gets maintained, improved, and tested on an ongoing basis.

If you are already running a ​ChatGPT Ads campaign, checking attribute completeness is the single highest-impact improvement you can make to ad performance.

The Full Performance Ad Stack Is Now In Place

ChatGPT's ad platform has moved from experiment to full performance advertising infrastructure. CPC bidding launched first. Self-serve access followed. Product feed ads arrived on May 12, 2026. CPA bidding is in active development. OpenAI hired a senior performance-advertising executive to lead the buildout.

This progression mirrors what Google and Meta did years ago: start with brand placements, then add performance formats, then build the self-serve and bidding layers that let mid-market advertisers run at scale. ChatGPT is compressing that timeline.

Here is how the two feed-based ad channels compare today:

The two channels are complementary, not competing. A shopper who searches Google for "hiking boots" and a shopper who asks ChatGPT "what boots should I get for a rainy weekend hike" are at different points in the buying process. The Google shopper is comparing options. The ChatGPT shopper is asking for a recommendation. Your feed serves both, but conversational placement rewards richer attribute data.

Retailers who already manage a Google Shopping feed are closer to ChatGPT readiness than they think. The feed format is the same. The gap is attribute depth, and closing it improves performance on both surfaces.

How To Prepare Your Catalog For ChatGPT Product Feed Ads

95% of merchants see AI agent traffic on their sites, but only 20% have machine-readable catalogs (May 10, 2026 panel data). Closing that readiness gap starts with five concrete steps.

• Step 1: Audit attribute gaps. Run a completeness check on your feed. Flag any product missing material, occasion, color, or finish attributes. Export your feed, count the empty fields per attribute column, and rank attributes by gap size. The attributes with the most blanks are your first priority.
• Step 2: Select your 100 strongest SKUs. Pick products with the best images, most complete attributes, and accurate pricing. These become your sample submission to OpenAI. Choose a mix that represents your catalog's breadth: different categories, price points, and product types. A sample of 100 nearly identical products will not demonstrate that your full catalog meets the quality bar.
• Step 3: Rewrite descriptions for conversational use. Descriptions should answer natural-language questions, not repeat keyword strings. "Waterproof leather hiking boot with Vibram sole, fits wide" works. "Boot hiking waterproof leather men's" does not. Read each description and ask: if a shopper asked a question about this product, would the description give the AI enough information to recommend it?
• Step 4: Install tracking. Add the ChatGPT Ads conversion pixel to your site before submitting your sample. Set up purchase and add-to-cart events. Without tracking in place, you will not be able to measure performance or optimize bids once your ads go live.
• Step 5: Connect through Ads Manager or a partner. Submit through OpenAI's Ads Manager Beta or through Criteo/StackAdapt if you already run with those platforms. If you use Criteo for your existing display or retail media campaigns, the feed integration is straightforward since they are the launch ad-tech partner.

Passionfruit's ​generative engine optimization service covers catalog readiness as part of a broader AI visibility program, so the feed work and the organic AI agent strategy are built together rather than treated as separate projects.

Make Your Catalog The Reason You Win, Not The Reason You Wait

The first move is a current-state audit. Check your feed's attribute completeness. Run a sample of your products through ChatGPT to see what comes back. Fix the gaps before submitting the 100-product sample.

The retailers who move first on product feed ads will have a compounding advantage: better data, earlier performance signals, and a head start on the CPA bidding format when it launches. Waiting until the channel is "proven" means competing against advertisers who already have months of conversion data to optimize from.

For brands that need the feed work and the AI visibility program built together, see how Passionfruit's ​generative engine optimization service runs catalog readiness alongside organic AI visibility, and ​talk to the team about where your feed stands today.