OpenAI is selling $60 cost-per-thousand-impression ad placements on ChatGPT. The tool to measure whether those ads work does not exist yet. It is being built.
Six weeks into running ads, OpenAI has recorded $100 million in annualized revenue, according to Reuters. It also quietly dropped its minimum spending threshold from $250,000 to $50,000 last week — signaling a push to broaden the advertiser base beyond the brand-awareness budgets that have carried the early test, per Digiday. But the structural problem has not changed: without conversion tracking, the company can sell impressions, not outcomes.
The ads manager, internally code-named "bazaar," runs on broader infrastructure OpenAI calls "tapestry," code reviewed by AdWeek shows. Snippets from the platform reveal support for click and conversion-based campaigns is listed as "coming soon" — even as the product today is limited to impression-based buying. The architecture suggests where OpenAI is headed. The product does not yet live there.
"Without knowing what types of conversions, it's unclear how this scales for performance marketing," said Juozas Kaziukėnas, an advisor at ecommerce ad platform Quartile and former Meta ad infrastructure engineer.
The business case rests on a premium that the measurement gap makes hard to justify. ChatGPT's $60 cost-per-thousand impressions is nearly double Google's display advertising rate, according to industry rate cards. But OpenAI offers no behavioral targeting, no audience size estimates, and no optimization tools — the standard toolkit for any performance marketer deciding where to put budget, per Digiday's review of the platform. The ads manager is updated daily and has A/B testing infrastructure in place, according to people familiar with the platform. But user profile-building for targeting, conversion tracking, and advanced analytics are all still missing.
OpenAI's financial exposure makes the commercial urgency clear. The company is projected to lose $14 billion this year, per projections cited by Digiday. It has set a target of $102 billion in advertising revenue by 2030 — 36 percent of a projected $300 billion total revenue — according to The Information, which reported the forecast. To reach that, OpenAI would need to grow its ad business roughly 20 times faster than Netflix did at an equivalent stage, according to Nate Elliott, principal analyst at eMarketer. Netflix had years of viewership data and mature tracking infrastructure when it started scaling ad sales. OpenAI is building its measurement stack while simultaneously selling against it.
"The ads manager is just plumbing," said Robert Webster, founder of AI marketing consultancy TAU. "Right now advertisers don't know what a ChatGPT user is worth, and neither does OpenAI."
The conversion tracking pixel is load-bearing infrastructure for the business model. Brand-awareness spend has a natural ceiling — it does not compound. Performance budgets do. OpenAI needs the latter to have any shot at the 2030 target. But performance advertisers do not pay premium CPMs on faith. They pay when the data shows return. The tracking code is the proof. Without it, the $60 CPM is a price tag nobody can verify.
What to watch: when conversion tracking ships — and OpenAI has not committed to a date — the company will have to show actual return on ad spend, not just impression counts. If the numbers are good, the $102 billion target becomes thinkable. If they are not, the brand-awareness spend will plateau, and OpenAI will have to find another way to close the gap.
