A closer look at OpenAI’s ads manager – and how much work it still needs

The one exception was Google, where Adwords and its self-serve manager launched together in 2000. It went on to become one of the most profitable ad businesses ever built. 

OpenAI’s execs have clearly taken notes because the upside is hard to ignore. An ads manager turns an ad business from a series of direct deals into a self-serve marketplace — one where advertisers set budgets, target audiences and measure results without a salesperson in the room. That’s what it did for Google. The question is whether OpenAI can make the same case. 

For now, at least, it looks the part. The ads manager bears a passing resemblance to Google’s own, according to a video of the dashboard that Digiday reviewed. Whether the functionality follows is another matter. 

As it stands, advertisers can only pay per impression, meaning they’re charged each time an ad is shown, regardless of whether anyone clicks. Cost-per-click and cost-per-acquisition options — the models most online advertisers actually rely on — are listed as “coming soon”. Until they arrive, advertising in ChatGPT remains largely off-limits to performance advertisers, who make up the bulk of the industry. Targeting is equally basic: advertisers can feed in keywords or free-text hints and restrict by country, but there are no demographic targeting or audience buying tools. Reporting is thin too — it stretches only to impression and click graphs. There’s no audience size estimates or optimization tools. 

That’s the product today. Behind the scenes, it looks different. The platform is being updated daily, according to an ad exec who spoke on condition of anonymity. A/B testing infrastructure is in place, feature flags mean different advertisers see different versions of the product, and recent additions include bulk upload support and onboarding screens for new advertisers. 

“Building an ads manager isn’t just a UI problem,” said Shirley Marschall, an ad tech consultant. “You’re dealing with real-time bidding logic, measurement, attribution, fraud prevention and more. Getting all of that to work reliably at scale is non-trivial. Which makes the irony hard to miss: this might be one of the few products that’s simply too complex to vibe-code.”

Little wonder OpenAI has yet to commit to a hard launch. Managing ad tech at scale means assembling a lot of moving parts simultaneously — and some of the most critical are still missing. User profile-building for targeting, conversion tracking, advance reporting and analytics — none of it is in place yet. But OpenAI has made clear to partners, even at this early stage, that the ads manager isn’t peripheral. Without it, there’s no ads business. And without that, the path to profitability looks considerably harder. 

“They’re doing things backwards, and they’re doing things in hopes to accelerate the adoption of their advertising, especially among smaller advertisers,” said Juozas Kaziukėnas, independent e-commerce and AI analyst and advisor.

What’s remarkable is how quickly that urgency materialized. As recently as 2023, OpenAI CEO Sam Altman was publicly dismissive of advertising as a revenue model. Three years later, the company has a dedicated ads team, trading deals with agencies and a self-serve manager in testing. From disdain to infrastructure in under three years.

“OpenAI is launching its ads manager early because advertising is now an urgent necessity to fund its massive compute costs and growth,” said Karsten Weide, principal and chief analyst at W Media Research.

The numbers make that plain. 

Last year, OpenAI burned through $8 billion in cash. This year that figure is projected to hit $25 billion, rising to $57 billion the year after. The losses are expected to continue until 2030 when the company projects it will turn cash flow positive, generating $40 billion in free cash. Advertising is central to that turnaround. OpenAI’s own projections put it at 36% of total revenue by then, or $102 billion of projected $300 billion total.

“OpenAI can’t reach $100 billion in annual advertising revenue unless it serves the SMB market, and doing so requires self-serve capabilities,” said Eric Seufert, analyst at Heracles Media. “And ads manager functionality is mostly a solved problem with a straightforward design standard, so rolling it out now allows OpenAI to tailor its ad platform to the customer profiles that it needs to service in order to achieve its revenue goals (SMBs).”

To do that, it doesn’t just need an ads manager. It needs one fast — and it needs to build everything around it simultaneously. Which formats generate the most revenue without alienating users? Which targeting signals are available and how to make them actionable? How to measure performance in a way advertisers actually trust. How to extend that across every market? And how to make the case, ultimately, that advertising in ChatGPT works better than anywhere else?

“Even then, it would be a staggering achievement,” said Nate Elliott, principal analyst at eMarketer.“It’s worth comparing them to Netflix, which also has almost a billion users and recently introduced advertising. Last year — their third year selling ads — Netflix made $1.5 billion from advertising, far exceeding early projections. But to hit its target, OpenAI would need to grow its ad business about 20x faster than Netflix has. And unlike Netflix, OpenAI doesn’t have access to a proven, high-value ad format to lean on.”

None of that makes success impossible. OpenAI has assembled a leadership team that knows how to build an ads business and it has a meaningful head start over its most obvious rival in the AI chatbot space — Google’s Gemini, which has yet to make a serious move on advertising. The ads manager needs work. A lot has to go right. But the pieces are being assembled, faster than the precedents would suggest is normal. That, in itself, is the story. 

“OpenAI is closer to Meta’s starting position than Google’s,”  said Robert Webster, founder of AI marketing consultancy TAU. “When Google launched, natural search volumes were already visible, so advertisers could price the opportunity from day one. Meta took years to find its value prop. Right now advertisers don’t know what a ChatGPT user is worth and what the right strategy is, something neither they or OpenAI know today. The pilot is both finding out. The ads manager is just plumbing.”

OpenAI did not respond to Digiday’s request for comment.