OpenAI wants advertisers to pay Google-level prices for clicks inside ChatGPT. It does not yet know if the signals behind those clicks are worth it.
The company launched cost-per-click advertising this week at $3 to $5 per click, The Information reported Tuesday, upending a CPM-only model that had been the only option since the pilot began in early February. The shift is a direct play for performance advertisers, the buyers who control the bulk of online ad budgets and who will not touch impressions alone. But it also exposes the central question OpenAI has not yet answered: when a user clicks an ad inside a ChatGPT conversation, does that click carry the same buying intent as a Google Search click, or does it behave like a Meta click, where users are browsing rather than buying?
Meta CPCs run three to five times cheaper than Google Search, according to digital ad agency Adthena, not because Meta's inventory is worse but because intent behind the clicks differs. Social users browse. Search users are looking for something specific. ChatGPT sits somewhere between: a conversation that may or may not reveal commercial intent. The $3 to $5 CPC asks advertisers to bet on the former while the market is still testing the latter.
One signal the market has already sent: CPMs, the cost to reach 1,000 users, have dropped from $60 at launch to as low as $25, a decline that forced OpenAI to cut its minimum ad spend commitment by 80 percent, from $250,000 to $50,000, Digiday reported. That contraction is the market's verdict on the unproven signal inside ChatGPT, delivered before OpenAI has produced the measurement infrastructure to contest it.
Nicole Greene, vice president analyst at Gartner, said the CPC model does not solve this problem. It just moves the pricing question downstream. The real question is whether a click inside ChatGPT behaves like a search click or a social click. Until there is performance data that answers that, advertisers are being asked to pay search prices for an unproven signal.
OpenAI has told investors it expects $2.4 billion in advertising revenue this year, growing to $11 billion in 2027 and $100 billion by 2030, according to investor presentations reported by Reuters and confirmed by multiple outlets. Those are the projections. The measurement reality is still being built.
Sensor Tower data from the first 30 days of the pilot shows retail brands accounting for roughly 44 percent of all identified advertisers, with more than 100 individual brand promotions recorded, according to AlmCorp. The roster includes Target, which has described its ads appearing based on keyword matches in user prompts. Someone asking ChatGPT for countertop appliance recommendations might see a Target ad. The ad appears below ChatGPT's organic answer, clearly labeled as sponsored, with no influence over the answer itself, according to OpenAI's public specifications.
WPP, Omnicom, and Dentsu are among the agencies that committed to the pilot, with some brands dedicating $200,000 to $250,000 for initial tests, CNBC reported. That is double a typical experimental commitment, according to people familiar with the agreements. Some advertisers told CNBC the slow rollout of the ad program meant their full budget commitments were unlikely to be spent by the end of the pilot period in March. OpenAI said the cautious rollout was intentional. Dentsu said it set realistic expectations and was seeing volume increase week-over-week.
Claire Holubowskyj, senior research analyst at Enders Analysis, said the CPC model is OpenAI's answer to declining CPMs and the need to keep advertiser demand growing. CPC gives OpenAI a way to grow ad revenue that does not depend on holding CPMs up, she told Digiday.
OpenAI is racing to build the infrastructure to answer whether the CPC bet pays off. The company posted a job listing this week for its first advertising marketing science leader. The role will own advertiser measurement strategy from the ground up, building reporting that connects to attribution models, incrementality testing, media mix modeling, and geo experiments. The person will work with third-party measurement partners and clean-room providers, and will be the external face of OpenAI's measurement methodology to advertisers. The job listing gives a sense of how early this is: the person starts as an individual contributor before scaling a team.
If the clicks perform, OpenAI has a credible case against Google's search ad monopoly. If they do not, the CPC model becomes a pricing structure built on a question the company has not yet answered.
OpenAI did not respond to multiple requests for comment on this story.
