ChatGPT Ads for Ecommerce: The Format Breakdown, the Real Constraints, and How to Actually Run It in 2026

Last updated: June 25, 2026. The ChatGPT Ads platform is evolving rapidly. Flag for quarterly review — re-verify all format specs, targeting options, and partner availability against ads.openai.com before publishing or acting on anything in this article.

Over 2,000 brands are already running ads inside ChatGPT. That’s not a projection — that’s what Criteo, OpenAI’s first official ad-tech partner, reported as of mid-2026. Most of those brands are ecommerce players, and they got in early because the format OpenAI built for shopping isn’t a banner slapped inside a chat window. It’s something genuinely different.

That said, this channel is still in active testing. Some things that sound exciting are confirmed and live. Others are policy-announced but not confirmed shipping. We’ll be explicit about which is which — because conflating the two is how advertisers make bad budget decisions.

What the ChatGPT Ecommerce Ad Formats Actually Look Like

The format that matters most for ecommerce right now is OpenAI’s dedicated shopping ad unit. It comes in two orientations — portrait and landscape — and it pulls in structured shopping data: product price and customer reviews alongside your creative.

Think about what that means for intent. A user asks ChatGPT “what’s the best lightweight hiking boot under $150?” and your ad surfaces with a real price and a star rating baked in. That’s not a display impression. That’s a product card sitting inside a high-intent conversation.

The newer iteration of the format goes further: a larger hero image paired with dynamic CTA buttons that adapt based on context — “Shop Now,” “Buy Now,” or “Learn More.” The button that appears isn’t static; it’s matched to where the user seems to be in their decision process based on the conversation. That’s meaningful differentiation from anything in Google Shopping or Meta’s catalog ads right now.

What we don’t know yet: exact dimension specs, how shopping data is ingested (feed-based, manual, or scraped), and how much control advertisers have over which CTA renders. If you’re planning creative, build for flexibility — assume you’re not choosing the button text yourself.

How Targeting Actually Works (And Why It’s Not What You’re Used To)

If you come into ChatGPT Ads expecting to build keyword lists and audience segments the way you do in Google, you’ll be confused within an hour. The targeting model here is purely contextual.

OpenAI uses the conversation topic and conversation history to determine ad relevance. There are no keyword targets, no intent-based audience lists, no demographic overlays. The system matches your ads to conversations where your product category is genuinely relevant based on what the user is actually talking about.

For ecommerce, this is actually a reasonable trade-off. A user mid-conversation about upgrading their home office setup is a high-quality match for ergonomic furniture or monitor accessories. You don’t need to bid on “best ergonomic chair” as a keyword — the conversation context does that work for you.

Where it gets harder: you can’t exclude irrelevant contexts with the same precision you’d use a negative keyword list in Google. Your main lever for waste control right now is creative specificity — the more precisely your ad speaks to a specific use case or product category, the more naturally it fits (or doesn’t fit) a given conversation context.

Our full breakdown of how this targeting model compares to what you’re probably used to is in our piece on ChatGPT Ads targeting options — worth reading before you build your first campaign structure.

Bidding: What’s Real and What’s Not Available Yet

Current bidding options are CPC and CPM. There is no minimum spend requirement, which is why SMBs can access this through Criteo GO without a managed deal.

What’s not available: outcome-based bidding. No ROAS targets. No CPA targets. If you’re an ecommerce operator who runs Google Shopping with a tROAS strategy, this will feel like flying without instruments. You’re optimizing toward clicks or impressions, not revenue. That’s a meaningful constraint and you should budget accordingly — treat early spend as data collection, not performance delivery.

The absence of tROAS/CPA bidding also means you can’t let the algorithm optimize toward purchase events the way Smart Bidding does in Google. Conversion tracking matters enormously here, and the gap between brands who have their pixel and Conversions API set up correctly and those who don’t will be huge when outcome-based bidding eventually ships. If you haven’t sorted your measurement setup yet, our piece on ChatGPT Ads conversion tracking is the right place to start.

The Criteo Partnership: What It Actually Unlocks for Ecommerce Brands

Criteo was announced as OpenAI’s first official ad-tech partner, and this matters for ecommerce specifically because Criteo’s entire product DNA is retail and commerce. They built their business on dynamic product retargeting. The fact that they’re the first integration partner is not a coincidence.

For SMBs, Criteo GO is the self-serve access point. You don’t need to negotiate a direct OpenAI deal. If you’re already on Criteo’s platform for retargeting, the path to ChatGPT ad inventory may be shorter than you think — ask your Criteo rep directly about ChatGPT inventory availability in your region and category.

For larger ecommerce brands, the Criteo integration suggests the likely direction of the platform: commerce-first, catalog-aware, potentially feed-driven. Brands that already maintain clean product feeds (structured data, accurate pricing, review signals) will have a structural advantage when the format matures and more data surfaces inside the ad unit.

That said: verify current availability directly with Criteo before committing budget. The partner ecosystem around ChatGPT Ads is moving fast and what’s available in one market or vertical may not be available in yours.

Audience Tools and AI Creative: What’s Been Announced vs. What’s Live

In approximately mid-June 2026, OpenAI published “Ad Tools Terms” that define two new capabilities: first-party Audience Tools and AI Creative Tools.

The Audience Tools definition suggests the ability to use first-party data (your customer lists, purchase history, etc.) to inform targeting — a major step beyond pure contextual. The AI Creative Tools suggest OpenAI-assisted ad creative generation within the platform.

Here’s the important caveat: as of late June 2026, these were policy-announced and not confirmed live. We’ve seen enough platforms publish capability terms weeks or months before actual rollout to know that “terms exist” ≠ “feature ships.” Do not build your Q3 strategy around these capabilities. Watch ads.openai.com and Criteo’s release notes. When they’re confirmed live, they will substantially change how ecommerce targeting on this platform works.

When first-party Audience Tools do ship, ecommerce brands with clean CRM data will have a real edge. Start auditing your first-party data now so you’re ready to activate, not scrambling to clean lists after the fact.

Restricted Verticals: Know Before You Plan

The current test has explicit restrictions. Health, financial services, dating, and politics are restricted verticals. If your ecommerce business touches any of these categories — supplements, health devices, financial products — you may find your ads disapproved or your account flagged before you spend a dollar.

This isn’t unusual for a new platform. Google has similar category restrictions, and healthcare advertisers on Google already know how much friction restricted verticals create. We cover that in depth in our Google Ads for healthcare guide if you want a sense of how these restriction frameworks typically evolve.

If you’re in an unrestricted ecommerce category — apparel, home goods, sporting goods, consumer electronics, beauty (non-health-claim), pet products — you have a clean path. The early-mover advantage is real when the platform is only 2,000 brands deep and growing.

How to Actually Structure Your First ChatGPT Ecommerce Campaign

Given everything above — contextual targeting, no tROAS, early-format creative — here’s how we’d approach a first campaign:

1. Pick one product category, not your whole catalog. Contextual matching works best when your ad creative is tightly aligned to a specific use case. A running shoe ad will match better to running-related conversations than a “shop our full footwear collection” message. Narrow creative = better contextual fit.

2. Build for the product card format. Lead with a strong hero image that reads clearly at smaller sizes. Include your real price in the copy — the format surfaces price data, and if your ad copy contradicts or ignores what the unit is pulling in, you’ll create friction. Customer review signals matter here too; if your product reviews are thin, prioritize categories where your review volume is strongest.

3. Test both CTA framings. Even if you can’t control which dynamic CTA button appears, you can frame your ad copy to align with different stages of intent. “Compare all options” copy aligns with a “Learn More” moment. “Get it before it sells out” aligns with “Buy Now.” Build variants for both and see which earns better downstream conversion signals.

4. Set a measurement foundation before you spend. Without tROAS bidding, you need your own measurement setup doing the heavy lifting. OAIQ pixel plus Conversions API, UTM parameters on every destination URL, and a clear attribution window defined before launch. Budget without measurement is just a donation.

5. Treat month one as a data sprint, not a performance quarter. You’re learning contextual match quality, CTR by product category, and post-click behavior. Set that expectation internally so you’re not defending CPCs to a boss who expected Google Shopping ROAS from a brand-new channel in week three. If you’ve been through this conversation with Google campaigns before, our ecommerce campaign structure guide has a useful framework for setting realistic performance timelines that translates well here.

Frequently Asked Questions

Can I run ChatGPT product ads without a Criteo account?

As of late June 2026, Criteo is OpenAI’s first confirmed ad-tech partner and the main access point for both managed and self-serve (Criteo GO) advertisers. Direct access through ads.openai.com may become available — check there for updates. But if you want to move now, Criteo is your fastest path.

How is ChatGPT ecommerce advertising different from Google Shopping?

Google Shopping targets based on search keywords and product feed matching. ChatGPT’s ecommerce ad format targets based on conversation context — what the user is actively discussing. There’s no keyword bidding, no search term report, and no auction tied to a specific query. The intent signal is conversational, not transactional. That’s a fundamentally different model, and it means your creative has to work harder to signal relevance.

Is there a minimum budget to get started with ChatGPT Ads for ecommerce?

No confirmed minimum spend. CPC and CPM bidding are available with no stated floor. Through Criteo GO for SMBs, the barrier to entry is low. That said, you’ll want enough budget to generate statistically meaningful click and conversion data — we’d suggest treating anything under a few hundred dollars as directional signal only.

Can I target specific audiences with my ChatGPT shopping ads?

Not yet in the traditional sense. Current targeting is contextual — the platform matches your ad to relevant conversations, but you cannot upload customer lists, set demographic overlays, or build remarketing audiences (as of late June 2026). OpenAI’s “Audience Tools” were defined in their Ad Tools Terms but are not confirmed live. Watch for updates before building targeting strategies around those capabilities.

What product categories work best for ChatGPT ecommerce advertising right now?

Categories that benefit from considered purchase conversations — where users are researching, comparing, or asking for recommendations before buying. Think consumer electronics, home goods, outdoor/sporting equipment, apparel for specific use cases, pet products, and beauty. High-consideration, mid-to-premium price point products that benefit from appearing during an active “help me decide” conversation.

Are ROAS targets available for ChatGPT Ads?

No. Outcome-based bidding (tROAS, tCPA) is not available in the current test. You’re bidding CPC or CPM. This is a significant operational difference from running Google Shopping with Smart Bidding, and it changes how you measure and optimize performance — at least until OpenAI ships outcome-based bidding.

My brand is in health/wellness ecommerce. Can I advertise on ChatGPT?

Health is a restricted vertical in the current test phase. Wellness products that make health claims will likely face disapprovals. Pure lifestyle/fitness products without health claims may have a path, but expect extra scrutiny. Verify directly with your Criteo account team or through ads.openai.com before building a campaign.

Is This Worth Your Budget Right Now?

If you sell physical products in an unrestricted category, your measurement setup is solid, and you have budget to treat month one as a learning sprint — yes. The ecommerce-specific format with price data, review signals, and dynamic CTAs is the most commerce-native ad unit OpenAI has shipped. The contextual intent signal — catching a user mid-conversation about exactly what you sell — is genuinely valuable.

If you need tROAS-based efficiency out of the gate, or if your vertical is restricted, or if your product feed and first-party data are a mess — wait. Come back in Q4 when outcome-based bidding may be live and the Audience Tools situation is clearer.

The brands who’ll win here aren’t the ones who throw the most budget at it. They’re the ones who get their measurement right, start narrow on creative and category, and learn the contextual matching behavior before they scale. That’s how early-mover advantage actually works — not by spending more, but by learning faster.

If you want a second opinion on whether ChatGPT Ads fit your specific ecommerce situation — what budget makes sense, how to structure the test, and whether your current measurement setup can handle a new channel — our fit-test framework is a good starting point. And if you’re already running Google Ads for ecommerce and want to see how the two channels should work together rather than compete, we can walk through that too.

The window where 2,000 brands have this channel and millions don’t won’t be open indefinitely.