ChatGPT Ads via Criteo: A CFO-Ready Pilot Framework

ChatGPT Ads via Criteo: A CFO-Ready Pilot Framework

Three months ago, Criteo became the first adtech partner to plug into OpenAI's ChatGPT advertising pilot. As of Criteo's May 5 update, more than one thousand brands are now running live campaigns through that integration. The question for marketing leaders is not whether to pay attention, but how to structure a test that Finance will actually sign off on.

The conversion signal is the headline number. Criteo's initial data from 500 U.S. retailers showed users referred from LLM platforms converting at roughly 1.5x the rate of other referral channels. More recent reporting suggests that figure is approaching 2x in several retail categories. That lift matters because it changes the math on what you can afford to pay per click or impression. A channel with higher downstream conversion can absorb a higher CPM and still hit your CAC payback target.

The Mechanics of the Integration

Criteo's role is distribution infrastructure. Brands already using Criteo's commerce media platform can activate ChatGPT as an additional surface without building a separate workflow. Campaign budgeting, bidding, and creative run through the same console. OpenAI controls delivery decisions; Criteo handles demand aggregation and cross-channel optimization.

This matters operationally. OpenAI's May 5 announcement introduced a beta self-serve Ads Manager alongside CPC bidding and a Conversions API. But for teams already running Criteo retargeting at scale, the integration removes the friction of onboarding a new platform. You get ChatGPT inventory inside your existing media planning workflow, with the same reporting cadence and optimization levers you already use.

The agency dimension is also worth noting. Criteo's update highlights traction with both holding companies and independents. Tinuiti, for example, described "immediate interest from clients eager to explore the intersection of generative AI and commerce." That agency adoption accelerates access for brands that would not otherwise have direct relationships with OpenAI's sales team.

What the Conversion Data Actually Tells You

The 1.5x to 2x conversion lift is directionally useful, but it requires context. The sample is U.S. retailers observed over a short window. The comparison baseline is "other referral channels," which lumps together organic search, paid social, email, and affiliate. A 1.5x lift against that blended average is not the same as a 1.5x lift against your best-performing prospecting channel.

First Page Sage's 2026 report on ChatGPT conversion rates offers more granular data. Their analysis of 160+ companies found that ChatGPT-referred traffic consistently converts at higher rates than traditional SEO, with the gap widest in complex B2B categories like commercial insurance and pharmaceuticals. Simpler B2C categories like ecommerce and food and beverage showed smaller improvements.

The implication: if your sales cycle is long and your product requires explanation, the intent signal from a ChatGPT conversation may be more valuable than a keyword match. If you sell commodity products with short consideration windows, the lift may be marginal.

The Measurement Gap

Here is where the CFO conversation gets harder. OpenAI's measurement architecture is still maturing. Advertisers receive aggregate impressions and clicks, but the closed-loop attribution that performance marketers expect from Google or Meta does not exist in the same form. OpenAI has introduced a pixel and a Conversions API, but the attribution window defaults to seven-day click and one-day view, and the data advertisers receive is aggregated and non-identifying.

This is a structural choice, not a temporary limitation. OpenAI has been explicit that conversations are not shared with advertisers. That privacy posture is a feature for users and a constraint for measurement. You will need to build your own incrementality framework rather than relying on platform-reported ROAS.

"The real question for marketers is not whether ChatGPT ads will attract early spending, but whether they will become a durable part of the performance media mix. History suggests that the answer will depend less on audience and engagement and more on measurement architecture."

Brian Quinn, AppsFlyer

A Pilot Design That Finance Can Approve

Given the conversion signal and the measurement constraints, here is how I would structure a two-week test:

Start with a budget that is large enough to generate statistical significance but small enough to treat as a learning investment. For most mid-market commerce brands, that means $15,000 to $25,000 in spend. Run the test through Criteo if you are already on the platform; the operational lift is lower and you get cross-channel comparison data automatically.

Define your success metric before you launch. If you are optimizing for CAC payback, model the allowable cost per acquisition at your current conversion rate, then calculate what that number becomes at 1.5x conversion. If the CPM math works at the lower bound, you have a defensible case for continued investment.

Build a holdout. The only way to measure incrementality in a channel with limited attribution is to compare exposed and unexposed cohorts. If you are running Criteo retargeting across multiple surfaces, hold out a geographic or audience segment from ChatGPT activation and compare downstream revenue.

Instrument your landing pages. Since ChatGPT does not pass standard UTM parameters through its search results, use dedicated landing pages for ChatGPT traffic. Combine that with first-party survey data ("how did you hear about us?") and referral source monitoring in your analytics platform.

The Budget Allocation Question

Criteo's update notes that budgets are incremental, with marketers treating ChatGPT as an additive discovery channel rather than reallocating from existing channels. That is the right framing for now. The channel is too new and the measurement too immature to justify pulling dollars from proven performers.

The exception: if you are already running Criteo retargeting and your prospecting channels are hitting diminishing returns, ChatGPT activation through the same platform is a low-friction way to test a new surface without adding vendor complexity. The incremental cost is the media spend itself, not a new integration or reporting workflow.

Axios reported that OpenAI is targeting $2.5 billion in ad revenue this year and $100 billion by 2030. Those numbers suggest the platform will continue to invest in measurement and targeting capabilities. The question is whether you want to build institutional knowledge now, while CPMs are still finding their level, or wait until the channel is mature and competitive.

What This Means for Your Board Deck

If you are presenting a Q3 media plan to your board, here is the framing I would use: ChatGPT advertising is a high-intent discovery channel with early conversion signals that justify a structured pilot. The Criteo integration reduces operational friction for brands already on the platform. Measurement is immature, so the pilot should be designed around incrementality testing rather than platform-reported ROAS.

The risk is spending without learning. The opportunity is building a playbook for a channel that may become a meaningful part of the commerce media mix within 18 months. Model the downside, cap the spend, and treat the first test as a learning investment with a clear decision point at the end.

Finance will ask what happens if the conversion lift does not hold. The answer is that you will have spent a bounded amount to learn that the channel does not work for your category, which is itself valuable information. The alternative is waiting until your competitors have already figured it out.