Key Takeaways
- OpenAI now lets retailers connect product catalogs to ChatGPT and auto-generate ads from product names, images, and attributes, removing the manual setup burden
- Feeds mirror the structured files used for Google Shopping, supporting up to 1 million SKUs per advertiser and dramatically lowering the barrier to entry
- The move signals OpenAI's pivot from transaction-fee monetization toward capturing ad budgets currently spent on Amazon and Meta
OpenAI Turns Product Catalogs into Automated Ads Inside ChatGPT
OpenAI is turning product catalogues into automated ads, making it easier for retailers to scale campaigns.
searchengineland.comOn May 12, 2026, OpenAI rolled out a new "product feed" ad format inside ChatGPT, reported simultaneously by Search Engine Land and Digiday. The setup will be immediately familiar to anyone running shopping campaigns on Google: retailers connect their product catalog, set filters defining which products are eligible, and let ChatGPT automatically generate ads using product names, images, prices, and attributes from the feed.
From the user's perspective, nothing changes. Ads still appear below ChatGPT's response, clearly labeled as sponsored, and don't influence the response itself. What changes is what's happening behind the scenes — workflows that previously required building campaigns manually, product by product, are now driven by feeds. For brands with thousands or tens of thousands of SKUs, that's the difference between running ads on ChatGPT being theoretical versus actually viable.
Why Feed Support Is the Real Inflection Point
ChatGPT advertising launched in February 2026 with Criteo as its first ad tech partner, but the earliest implementations were heavily manual. Large retailers with deep catalogs couldn't justify the operational overhead of building campaigns one SKU at a time, which effectively excluded any brand with more than a few hundred products.
The new feed format removes that scalability ceiling. According to Digiday, the platform can handle up to 1 million SKUs per advertiser, with OpenAI currently asking new e-commerce partners to submit a sample of 100 products before granting access to full catalogs. The threshold opens the door to apparel, electronics, home goods, and furniture categories where inventory routinely runs into the tens of thousands.
The critical detail: the feed file required is essentially the same structured file retailers already submit to Google Shopping. StackAdapt CTO Yang Han, whose firm is one of the ad tech partners selling ChatGPT inventory, summed it up by saying that once OpenAI supports the format, his platform's feeds can pass through unchanged. For any retailer that has built a healthy Google Merchant Center setup, entry to ChatGPT advertising now carries essentially zero additional preparation cost.
OpenAI's Monetization Pivot Comes Into Focus
This update is bigger than a new ad format. It's a clear signal about how OpenAI plans to make money from commerce going forward.
Until recently, OpenAI positioned Instant Checkout — letting users complete purchases inside ChatGPT — and the underlying Agentic Commerce Protocol (ACP) as its primary commerce monetization path, taking a small cut of every transaction. That bet has effectively been wound down in recent weeks.
Adding automations to generate ads from a product catalog is essentially table stakes in the age of AI. What OpenAI is doing is somewhat similar to what Google, Meta and Amazon already offer retailers. The difference is that ChatGPT is serving ads based on conversational intent rather than signals from search behavior, social engagement or browsing in a marketplace environment.
The product feed launch suggests OpenAI hasn't abandoned commerce as a category so much as it has changed the path it plans to monetize. Rather than taking a cut of transactions, OpenAI is going after the advertising budgets e-commerce brands already spend on Amazon and Meta. Search Engine Land frames the move as a direct play for ad budgets currently flowing to those platforms.
OpenAI has been laying the performance-advertising groundwork for months. It launched cost-per-click bidding earlier this year, introduced conversion tracking, and is reportedly developing CPA (cost-per-action) bidding. Hiring David Dugan, a Meta veteran and one of the architects of that company's performance ad business, as global head of ads reinforces the direction.
Can "Conversational Intent" Actually Convert?
Lower friction is one thing. The harder question is whether conversational intent can drive conversions at a level comparable to traditional search or marketplace signals.
That's the differentiator Search Engine Land highlights. Conversations with ChatGPT carry context that search queries rarely surface — for example, "I'm looking for an understated piece of jewelry for my wife's anniversary" includes situational and emotional cues a Google search query usually doesn't. Criteo's February 2026 client data showed that users arriving via LLM platforms convert at roughly 1.5x the rate of other referral channels, an early signal that the intent quality may justify a premium.
Whether that uplift holds up at scale, and whether it is sufficient to redirect meaningful ad budgets, is what the next few quarters will determine. Programmatic plumbing, larger sample sizes, and a maturing measurement stack will all sharpen the picture.
Three Things E-Commerce Operators Should Do Now
With the product feed format, ChatGPT advertising shifts from "an experimental channel a few large brands test by hand" toward "a standard channel alongside Google Shopping." Three priorities matter most right now.
First, clean up your Google Merchant Center feed. Because ChatGPT essentially reuses the Google Shopping feed spec, the quality of your titles, descriptions, categories, images, pricing, and availability data directly determines the quality of the ads ChatGPT generates. If your Shopping campaigns underperform today, ChatGPT ads will too — fix the feed before fixing the channel.
Second, set up LLM-referral measurement now. Isolate traffic from chat.openai.com and chatgpt.com in GA4 and build a dashboard comparing conversion rate, AOV, and repeat purchase rate against other channels. Whether you can defend a higher CPM for ChatGPT depends on whether you can actually see how LLM-referred users behave on your site.
Third, optimize your product data for AI agents, not just SEO. ChatGPT ads depend on structured product data and AI-driven matching against conversational context. Keyword-stuffed product titles don't help here — context like use case, occasion, target persona, and explicit attribute tagging in the feed do. Brands that treat product data as input for AI agents, not just search crawlers, will surface more often.
Closing Thoughts
The product feed launch is a symbolic step in AI search advertising's transition from experiment to infrastructure. Reusing Google Shopping feeds removes the most common excuse for staying on the sidelines, opening the channel to mid-market retailers who couldn't justify the manual lift.
For OpenAI, this is a pragmatic rebound from the Instant Checkout retreat — replacing a difficult transaction-fee model with a more conventional advertising business. With CPC live and CPA in development, the next few quarters of conversion data will decide whether ChatGPT becomes a true peer to Google, Meta, and Amazon as a performance channel, or settles into a complementary niche.
The smart play for e-commerce operators isn't to rush large budgets at the platform. It's to get the prerequisites right — feed quality, referral measurement, and AI-ready product copy — so that when the dust settles, you can scale into a channel you already understand rather than scrambling to build the basics under competitive pressure.
