OpenAI’s ad business is four months old, and the company just took it to Cannes International Festival of Creativity, the industry’s biggest stage.
That choice is the story.
A young, unproven paid media product doesn’t need a credibility tour unless the goal is to make it look permanent, which is exactly what OpenAI did, gathering reporters on a quiet patio while Chief Revenue Officer Denise Dresser told them the company is in the advertising business now.
For marketers, the useful signal isn’t that ChatGPT has ads. It’s that a new intent-driven channel is being legitimized before it has matured, and the people testing it today will understand it well before their competitors do.
Why OpenAI Thinks the Ads Will Work
The pitch, a low-key briefing and NOT a keynote, came from David Dugan, OpenAI’s head of global advertising and, until recently, a 12-year Meta veteran who ran global clients and agencies.
His argument started with a claim about behavior: people don’t open ChatGPT to scroll, they open it with a job to be done, a problem to solve or a thing they’re trying to buy.
Dugan argued that is what makes the app a good place for an ad. It’s a pointed claim from someone who spent over a decade at the company that perfected the scroll.
The number underneath it is the one to remember. OpenAI says roughly 20% of ChatGPT queries carry direct commercial intent. One in five conversations is someone actively researching or comparing something they might pay for, a far higher concentration of purchase signal than a social feed delivers.
Dugan framed the ads as a new format rather than a copy of classic search, and stressed that OpenAI is rolling them out deliberately, market by market. He called the company fully dedicated to advertising and cast the revenue as a way to subsidize broader access to its tools.
He had some early proof to offer. Cross-out rates, the share of people who hit the X to dismiss an ad, have dropped by half since launch, and OpenAI says serving ads hasn’t measurably changed how people use the app.
The traction is concentrated in travel, retail, health and beauty and financial services, all research-heavy categories. Underneath, the plumbing is filling in fast: cost-per-click now makes up most of the spend, a self-service ads manager has launched in English-speaking markets and partners like Criteo, StackAdapt, Adobe and LiveRamp are lowering the barrier to entry.
Truly, the venue made the point that the metrics couldn’t. You don’t bring a four-month-old product to advertising’s most self-regarding event unless you want the agencies and brands in the room to start treating it as a channel.
A Pitch With Two Audiences
The Cannes trip was aimed at two audiences at once, and the second one explains the urgency of the first.
The Marketers
OpenAI’s framing is that advertising is moving from an attention economy to what its executives call an intelligence economy, where usefulness is the win condition rather than time spent. In practical terms, that means the company wants ChatGPT treated as a performance and intent channel living in the decision layer of a purchase, the middle of the funnel, where people compare options, rather than the top, where they’re just becoming aware. That’s a different job than an awareness buy, which is why the pitch is aimed squarely at agencies and brands rather than a general audience.
The Investors
OpenAI has told people it expects to reach $100 billion in revenue within a few years, a target that advertising is meant to help carry. Reporting points to a confidential SEC filing and a possible public debut as early as this fall. A credible ad business makes that revenue story far easier to tell, so the legitimacy built on a marketing stage is quietly doing double duty for the IPO.
What the Pitch Skips
The strategic story is clean, but the operational reality is messier, and a few things deserve a marketer’s skepticism:
- The channel is still mostly a black box, and measurement needs to mature before most marketers can prove its return.
- Targeting and minimum spends will keep shifting as OpenAI learns what works, so anything you test now is a moving target.
- The upper-funnel promise Dugan likes, reaching someone early in researching a family trip to the Alps before any commercial intent shows, is the hardest value to measure and the easiest to oversell.
- Bigger budgets follow proven performance, and the proof isn’t all there yet.
The Guardrails OpenAI Keeps Naming
Trust is the thing that could sink the effort, so the company keeps drawing bright lines: sponsored content stays separate from the chatbot’s answers, conversational data isn’t shared with advertisers, users keep control of their history and the company talks up long-term value over near-term revenue. Whether those lines hold under IPO pressure is the open question. For now, they’re the reason the channel could last instead of flaming out.
Before You Test Anything
The instinct with a channel this young is to wait until someone else proves it works, and that instinct deserves more respect here than the early-mover hype suggests.
The proof isn’t in yet. Measurement is thin, the rules change from quarter to quarter and OpenAI has a $1T IPO-sized reason to make the channel look more finished than it really is. That doesn’t make ChatGPT ads a bad bet, just an unproven one.
If your category runs on research and comparison, here’s how to stay close to it without getting ahead of the evidence:
- Observe before you buy. Open a free or Go-tier account and watch how sponsored placements show up against organic answers in your category. Track which competitors appear, how the ads read and how often they change. This costs nothing and builds a point of view you’ll need later.
- Audit the gaps that matter to you. Before any spend, get specific about what you can’t yet see: how conversions are attributed, what targeting controls actually exist, what reporting you’d receive and how minimum spends are trending. If the answers aren’t clear, that’s your finding.
- Keep any test small and time-boxed. If you do experiment, treat it as research, not a campaign. Set a budget you’d be comfortable losing, a single question you want answered and an end date, then judge it against that question alone.
- Watch the pitfalls OpenAI hasn’t solved for. Attribution inside a black-boxed channel is hard to verify independently, so you’re partly taking the platform’s word on how your spend performed. There’s also brand safety to consider, since no one has really tested how ads sit next to AI-generated answers. The upper-funnel value Dugan describes, reaching someone before they show commercial intent, is the easiest claim to oversell and the hardest to measure. And because minimum spends and targeting rules keep shifting, whatever you learn today may not hold next quarter.
Where You Go From Here
OpenAI came to Cannes to make ChatGPT ads look inevitable, and it mostly worked. The intent is real, the channel is early and the proof is still thin. That’s reason enough to watch closely, ask hard questions and run small tests while the stakes are low. We’re doing exactly that: auditing how the channel behaves, where it fits and what it can’t yet measure.
If you’re weighing whether ChatGPT ads belong in your mix, or want help designing a small test that actually answers a question worth asking, let’s talk. And if you want to go deeper on how AI is reshaping search and paid media, see how we think about it.
