OpenAI Is Done Borrowing Someone Else's Ad Stack
Ad TechMarch 24, 2026· 6 min read

OpenAI Is Done Borrowing Someone Else's Ad Stack

Zach El-AminBy Zach El-AminAI-GeneratedAnalysisAuto-published5 sources cited

OpenAI Is Done Borrowing Someone Else's Ad Stack

For a company that once treated advertising like a moral failing, OpenAI is moving fast.

As of this month, ChatGPT is showing ads to all free and Go-tier users in the United States. Sponsored placements appear below responses — clearly labeled, impossible to miss, and powered by Criteo, which has been pitching advertisers on $50,000–$100,000 minimum commitments to get in front of OpenAI's 910 million users. Reuters confirmed the expansion on March 21, following an initial test that launched in February.

But the real story isn't the ad rollout itself. It's what OpenAI is building underneath it.

According to Digiday, OpenAI is actively hiring a monetization infrastructure engineer, an engineering manager, a product designer, a senior manager for ad revenue accounting, and a trust and safety hire dedicated specifically to the ads product. All full-time, all San Francisco, all building a proprietary ad stack from zero. Compensation bands go as high as $385,000 — not the kind of money you spend on plumbing you plan to outsource forever.

This is OpenAI officially becoming an ad platform. And if you've been in this industry long enough, you've seen this movie before.

The Borrowed Stack Problem

Right now, OpenAI's ads business runs on infrastructure it doesn't own. Criteo is the primary interface for ad buying and targeting. The Trade Desk is reportedly in active talks to come in at scale. OpenAI, by contrast, has no ad server, no targeting framework, no measurement stack, and no agency relationships — at least not yet.

Leaning on outside vendors to bootstrap was the smart call. Criteo and The Trade Desk have something OpenAI desperately needs: existing advertiser pipelines, agency integrations, and measurement credibility. You can't launch an ad product cold and expect brands to wire money into a platform with no performance history. So OpenAI borrowed the credibility while it figured out what it was actually building.

"If OpenAI wants to serve ads to 910 million users and make money sooner than later, buying rather than building ad tech is definitely their smartest play," eMarketer principal analyst Nate Elliott told Digiday. But he was clear that's the short game. The long game — the one these job postings signal — is full stack ownership.

The Netflix Playbook, Executed Earlier

This is exactly what Netflix did. When it finally got serious about advertising in 2022, it partnered with Microsoft to get up and running fast. Microsoft had the infrastructure, the sales relationships, the ad tech. Netflix had the audience. It was a marriage of convenience that lasted roughly two years before Netflix started systematically pulling the stack in-house, building its own ad server and launching its own ad platform in 2024.

Walmart did the same — leaned on a patchwork of DSP partnerships before eventually building Walmart Connect into a real retail media network with first-party data at its core.

OpenAI is signaling that same intent from day one, which is actually more aggressive than either Netflix or Walmart. They're not pretending the vendor relationships are permanent. They're hiring the team to replace those vendors while those vendors are still actively selling on their behalf.

"Developing an AdTech stack takes forever," said Karsten Weide, principal analyst at W Media Research, per Digiday. "But OpenAI desperately needs revenue, that's why they will partner with one or more DSPs. They will split the revenue and buy time to market that way."

Translation: Criteo's window is open, but it has an expiration date stamped on it.

The Trust and Safety Tell

The most revealing hire on the list isn't the infrastructure engineer. It's the trust and safety role tied specifically to the ads product.

OpenAI knows that advertising inside an AI assistant isn't like advertising on a search results page or a social feed. The danger isn't just brand safety in the traditional sense — it's whether ads can distort the actual responses the model generates. That's a problem no DSP integration can solve for you. It requires proprietary controls built into the serving layer itself.

Fidji Simo, OpenAI's CEO of Applications, told Wired last year that brand safety, user experience, and the risk of ads corrupting model outputs are "design constraints from the start," not afterthoughts. The trust and safety hire is that principle becoming an org chart line item.

This is actually where OpenAI has a legitimate technical moat that Google, Meta, and the programmatic ecosystem can't easily replicate. If they build ad systems that are natively aware of the model's response quality — systems that can suppress an ad when it would compromise the answer, or weight placements based on conversation context rather than just click-through proxies — that's a genuinely new format. Not just a new surface.

What the Industry Should Be Paying Attention To

The ad tech world has been treating ChatGPT ads as a distribution opportunity — another surface to buy against. That framing is going to age badly.

OpenAI's own published ad philosophy says they don't optimize for time-spent, won't let ads distort responses, and will always offer a paid, ad-free tier. That sounds like the standard responsible-tech PR boilerplate until you notice they're hiring the engineering team to actually enforce those constraints at the infrastructure level.

If they execute, they're not building a worse version of Google. They're building something the RTB ecosystem fundamentally cannot replicate: an ad system where the creative and the content are one and the same, optimized for an answer rather than a click.

That's either the most valuable ad unit ever invented, or the most dangerous. Maybe both.

The old ad tech players have a narrow window to matter here — while OpenAI's stack is still under construction, while the job postings are still live, while the infrastructure engineer is still being interviewed. Once that team ships v1, the conversation changes.


Sources: Digiday — OpenAI is building the ad tech stack it's currently borrowing | Digiday — OpenAI is hiring engineers, not ad sellers, first | Reuters — OpenAI to introduce ads to all ChatGPT free and Go users | OpenAI — Our approach to advertising | MediaPost — OpenAI Addresses Monetization, Hires Ad Engineers

This article was AI-generated. Learn more about our editorial standards

Share:

Report an issue with this article