The Commerce-First Chatbot—The OpenAI & Criteo Strategic Alliance

Then came the more interesting development: Criteo became the first adtech partner integrated into the ChatGPT ad pilot. The pitch centres on commerce intelligence surfacing inside conversational discovery, where people actively seek recommendations rather than simply searching for products they already know they want.

What makes the combination significant is that it's one of the first real proofs that conversational AI can monetize without degrading the user experience. For brands and agencies, it also crystallizes something bigger: we're entering agentic commerce, where chat doesn't just inform—it shortlists, validates, and increasingly enables purchase.

TL;DR: Three takeaways to keep in mind

• Conversational intent is a different species of demand. In Criteo’s aggregated observations across 500 U.S. retailers (Feb 2026), users referred from LLM platforms like ChatGPT convert at ~1.5x the rate of other referral channels.
• Commerce intelligence is becoming the decision layer. Criteo’s thesis is straightforward: recommendation quality rises when you use structured commerce signals, not just product descriptions.
• The “recommendation” is the new shelf space. The winners will be the brands that make their product truth usable for AI systems and measurable for finance teams.

The pilot is small, but the signal is big

OpenAI’s stated principles read like a checklist of everything that has gone wrong in ad-funded platforms over the last decade: keep answers independent, keep conversations private from advertisers, clearly separate sponsored content, and give users control.

That framing isn’t PR fluff. It’s an admission that answer-led interfaces have a tighter tolerance for anything that feels like influence. OpenAI is already building guardrails around integrity and safety, including an “ads integrity” effort and “know your customer” style verification to reduce scam risk.

If you're a CMO, this should sharpen your focus. Ads in chat are a given. The question worth answering is what kind of ads can survive inside an interface people treat like an advisor.

Conversational intent isn’t funnel-stage theater

Too much media planning still runs on broad, comfortable abstractions—awareness, consideration, conversion—without interrogating what those stages actually look like now.

None of that vocabulary holds up in conversational discovery. People arrive with a request that already carries budget, constraints, preferences, and timing baked in. They're not browsing; they're handing over a brief.

‍

{{26-The-Commerce-First-Chatbot-1="/tables"}}

‍

A more accurate model looks like this:

Ask → shortlist → validate → act

• Ask: “What’s the best running shoe for marathon training if I overpronate?”
• Shortlist: The assistant narrows options based on constraints and known tradeoffs.
• Validate: Specs, reviews, fit guidance, price, availability, shipping, returns.
• Act: Click out, or increasingly, buy in place.

The 1.5x conversion lift starts to make sense in that light. There's no magic performance here—just what happens when the input is high-intent by design.

Why Criteo matters here

Criteo’s advantage has never been “it can serve an ad.” Plenty of platforms can do that. The advantage is commerce intelligence: signals derived from transactions, product catalogs, retailer relationships, and decisioning tuned for outcomes.

Criteo describes its footprint as 17,000 advertisers, more than $4B in annual media spend activation, and unique access to over $1T in annual commerce sales. That’s the kind of foundation you need if you want sponsored placements in a chat interface to feel like helpful options rather than interruptions.

It also sets up the real strategic shift: ChatGPT ads stop looking like an exotic experiment and start looking like a commerce channel with a familiar operating model.

The conversion premium is real, but it’s not the whole story

Yes, early indicators suggest LLM referrals can outperform other referral traffic. But that doesn’t mean you should treat ChatGPT like another performance faucet and turn the handle.

The real opportunity is upstream: influencing what makes the shortlist.

‍

{{26-The-Commerce-First-Chatbot-2="/tables"}}

‍

Criteo has been explicit that recommendation quality improves when the system has access to structured commerce signals. In internal testing (Jan 2026), Criteo reported up to a 60% improvement in recommendation relevancy versus approaches based only on product descriptions.

Call that what it is: an early warning to every brand with messy product data. If your catalog is inconsistent, your availability is unreliable, or your variant attributes are incomplete, you’re going to lose recommendation share, even if you “win” an impression.

The “VIP gate” and why it won’t last

Right now, access is constrained. Reports around the ad pilot have referenced high minimum commitments and premium pricing, with early coverage citing a $200,000 minimum and CPMs around $60. 

That’s not a long-term market structure. That’s a controlled test.

But it’s still telling that major holding companies are participating, even under those constraints. Adweek reported investment from Omnicom, WPP, and Dentsu, with Omnicom noting dozens of clients securing placements. 

Criteo’s integration is the piece that points toward scale. If conversational ads are going to become a durable channel, they need buying and measurement workflows that don’t require bespoke relationships for every experiment. This partnership is a step in that direction. 

Trust is the constraint in answer-led media

The market is already testing two paths.

As mentioned, OpenAI is building an ad model with explicit guardrails about answer independence and conversation privacy. At the same time, Perplexity has publicly moved away from advertising, arguing that even well-labeled ads can make users suspicious of the product’s objectivity.

You don’t have to pick a side in that debate to learn from it.

The practical implication for advertisers is straightforward. Anything that feels like it bends the answer will burn out fast. Being additive to the experience is the baseline requirement for showing up at all.

Looking ahead: agentic commerce makes product feeds a growth lever

The alliance lands at the same time OpenAI is publishing concrete plumbing for commerce inside ChatGPT. The Agentic Commerce Protocol and Instant Checkout are designed to let merchants provide product feeds, integrate checkout APIs, and enable purchases through the interface. 

The knock-on effect is a redefinition of "optimization." It expands well beyond creative and bid strategy into data structure, fulfilment truth, and user experience continuity.

It also accelerates a discipline many teams are already circling: generative engine optimization (GEO)—making sure your brand and products are represented accurately and usefully in AI-driven discovery. Even the academic community is formalizing this concept and the mechanics behind it.  

How to prepare now: a practical checklist

If you’re treating this like a 2026 media test, you’ll miss the bigger change. Treat it like an operating model shift.

1. Clean up product truth for machines, not just humans. Variant attributes, category taxonomy, sizing logic, shipping and returns, availability, and price consistency decide whether you’re recommendable. Instant Checkout and product feed requirements make that direction of travel explicit. 
2. Decide what you’re optimizing for. Impressions and clicks will show up first because they’re easy to instrument. Senior teams should define success in business terms: incremental demand, new-to-brand customers, contribution margin, and downstream repeat.
3. Separate “influence” from “harvest”. Conversational discovery is often an advisor moment. Retargeting might capture the final action, but it won’t tell you what created the shortlist. Build measurement that can isolate incremental impact.
4. Put governance around claims and safety. In chat, brand safety isn’t only about where the ad appears. It’s also about what the ad implies and how it relates to the user’s request. The integrity work OpenAI is doing is a hint at how sensitive this surface will be. 
5. Invest where transparency stays intact. As new surfaces emerge, the risk is drifting into black-box buying where you can’t explain why spend moved or why a product was favored. That’s where an open, accountable approach becomes a competitive advantage.

Where AI Digital fits in this new surface

Agentic commerce will reward teams that can move fast without losing control of measurement and governance.

Our perspective at AI Digital is rooted in three ideas:

• Open Garden: keep transparency and portability as new discovery environments grow, so strategy and data ownership don’t get trapped inside one platform’s rules.
• Smart Supply: select supply for quality and outcomes, especially as conversational inventory expands and the market experiments with new buying paths. 
• Elevate: unify planning, optimization, and insights across channels, with a workflow that helps teams ask better questions and act on the answers quickly.

The goal isn't to chase every new format. It's to build the foundation that lets you test what matters, measure what counts, and scale what holds up.

Closing thought: win the recommendation

The OpenAI–Criteo partnership doesn’t guarantee that conversational ads become the next mega-channel. It does something more valuable: it shows what a viable model could look like when commerce signals, decisioning, and trust guardrails work together.

If your team wants to take advantage of this shift, start with product truth and measurement design. The media plan will follow.

If you’d like to talk through how to structure tests, build governance, and measure incrementality as conversational commerce scales, reach out. I am always happy to talk!