OpenAI has added cost-per-action bidding to the ChatGPT ads manager, letting a pilot group of advertisers pay only when a user takes a defined action rather than per impression or click. The update, communicated in an email to early pilot advertisers and reported by Digiday, gives OpenAI its first performance-grade buying objective.

How CPA Differs from CPM and CPC

ChatGPT advertisers had been buying inventory two ways: cost-per-thousand-impressions, where they pay for every 1,000 views the ad receives regardless of user response; and cost-per-click, where they pay each time a user taps the ad regardless of what happens after the click. Both models put the result risk on the advertiser. A brand running a $10,000 CPM campaign at a $20 rate gets half a million impressions and no guarantee of a single click or conversion. CPC sharpens that exchange to the click level but still stops there.

CPA flips the equation. Under outcome-based bidding, the advertiser pays only when the impression or click leads to a result the brand defined in advance, such as a site visit, a registration, or a completed purchase. The platform absorbs the cost of any ad serve that does not produce that outcome and recovers the spend only on completed events. For the advertiser, that converts ad spend from a probabilistic pool into a price per result; for OpenAI, it requires the targeting and prediction quality to deliver outcomes profitably across a varied advertiser base.

The Infrastructure CPA Needed

That math only works when the platform can confirm whether an outcome happened. OpenAI added that capability in recent weeks with its pixel, a tracking tag that advertisers install on their own sites to register conversions back inside ChatGPT’s ads manager. Without the pixel, the platform had no way to attribute a downstream action to a specific ad impression, let alone charge against one. The CPA option is gated on having that pixel deployed and conversion events configured.

According to the email sent to pilot advertisers, “any account that has conversions set up by Monday, June 1st, will be granted early access by June 5th.” The gate is conversion-tracking configuration rather than spend tier or campaign type, which means an advertiser already on the platform needs the pixel firing this week to be in the first wave.

The Two-Month Build Inside the Ads Manager

CPA closes a buying-model arc the ChatGPT ads team built out across roughly two months. Earlier this year, OpenAI opened the self-serve ads manager to U.S. advertisers as its first general-availability buying surface, with pricing on CPM. It then shifted to CPC pricing to give brands a click-based response model. Around the same window, the team cut the minimum commitment fee to $50,000 to widen the advertiser pool and routed retailer product catalogs into automated shopping ads to lower the creative lift. The pixel followed. CPA is the layer that turns the rest of those moves into a coherent performance product.

Asad Awan, OpenAI’s head of monetization, signaled the move three weeks ago at a press briefing, describing outcome-and-conversion bidding as “in the plan” and “should be done soon.”

Performance Hires Arrive Alongside the Feature

OpenAI has been building out the operational side of its ads business in parallel. Dave Dugan, formerly Meta’s ads boss, now leads the OpenAI ads org. Archana Joshi, an ex-Meta executive, joined as a founding member of the Ads Go-To-Market Revenue and Strategy Operations team, a function focused on scaling advertiser partnerships. Sam Mulinder, previously Snap’s head of marketing science for North America and mid-market, joined to build OpenAI’s marketing science function from the ground up. The hiring pattern leans heavily on operators who built the performance machinery at the platforms ChatGPT now competes against.

Where ChatGPT Lands in the Performance Landscape

CPA bidding is the standard objective inside Meta Advantage+ and Google Performance Max, the two products that absorb the bulk of direct-response budgets in digital advertising. Adding the same buying model gives ChatGPT advertisers a like-for-like point of comparison they have not had before. Claire Holubowskyj, senior research analyst at Enders Analysis, told Digiday the launch “expands its offering while diversifying its pool of advertisers, and aligns its product more closely with that of Meta and Google.”

The timing matters. Meta is publicly courting AI-platform ad buying through new connectors that route ads into chat assistants, and Google has been folding more conversion bidding into AI Mode and Demand Gen. ChatGPT entering the performance category on familiar pricing terms gives advertisers a way to compare the three on the same scoreboard.

HOME
Related News