If you want to shape media strategy—not just optimize it—upgrade your expertise. The ADWEEK MiniMBA in Marketing equips you to lead with confidence and credibility. Register now.
StackAdapt is quietly courting advertisers to test ads inside ChatGPT. The independent demand-side platform is dangling CPMs as low as $15 alongside discounted platform and management fees. The company is framing the push as early access to a new “discovery layer”—one that captures people in the middle of researching and comparing products on ChatGPT.
According to a pitch deck titled “OpenAI x StackAdapt Limited Pilot Program,” shared with select buyers on March 27 and reviewed by ADWEEK, the company is positioning the offering as an early-stage test inside a still-developing ad system.
“StackAdapt has partnered with OpenAl to enable advertising within ChatGPT, one of the fastest growing consumer platforms in the world,” the deck reads.
StackAdapt describes that system as a “proto-auction model.” Unlike a traditional bid-based exchange, pricing fluctuates based on how many advertisers are competing for the same user prompt.
The deck outlines CPMs starting at $15 for highly niche placements where only one advertiser matches a user’s prompt. But when two or more advertisers are eligible, prices can jump to as high as $60 CPM.
“It’s very difficult to justify a $60 CPM on a campaign where you’re trying to narrow down on ROAS or CPA,” according to an agency executive partaking in the pilot. They added that the upper end far exceeds typical programmatic benchmarks.
The pilot itself carries a $50,000 minimum spend and is slated to run through the end of May, according to the agency executive. That is broadly in line with other OpenAI partners such as Criteo, which is offering a similar product.
Where Criteo is leaning into performance, StackAdapt is pitching presence—encouraging brands to be part of the conversation as users engage with AI assistants during research and consideration phases, the executive said.
That approach also marks a shift from OpenAI’s earlier direct deals, which until recently carried minimum spends as high as $200,000, ADWEEK previously reported. Those thresholds have since come down with recent deals ranging between $100,000 and $150,000, the executive said. “As part of the pilot, we have adjusted the minimum spend as we learn from early use and feedback,” an OpenAI spokesperson told ADWEEK. They stopped short of confirming the figures.
StackAdapt’s pitch lands as competition heats up among ad tech players to broker access to ChatGPT inventory, The Trade Desk is also reportedly in talks with OpenAI. While the opportunity promises proximity to high-intent queries, it also exposes a new set of trade-offs—limited targeting, volatile pricing and measurement gaps that advertisers are still trying to navigate.
An OpenAI spokesperson told ADWEEK that “StackAdapt has the same pricing as every other buying channel in the pilot.” The minimum bid remains $60, the spokesperson added, but delivery can clear at a lower effective CPM in some cases. Despite the deck, the company clarified “advertisers are not able to select a $15 CPM for this inventory.”
For OpenAI, which is preparing for a highly anticipated IPO, the emerging ad model signals a potential scaling challenge where mostly large-spend advertisers can afford rates as high as $60 CPM—matching top-tier television inventory including Sunday Night Football. Axios has reported that OpenAI is targeting $2.5 billion in ad revenue this year with ambitions of reaching $100 billion by 2030. OpenAI recently expanded its ads pilot to Canada, Australia, and New Zealand.
“If ChatGPT wants to get anywhere close to its 2030 ad revenue targets, its ads must be priced to sell, and right now, even at the lowest number ($15 CPM), are over 50% higher than the average CPM we estimate for Meta in the current quarter,” said Max Willens, principal analyst at Emarketer. “If ChatGPT wants to scale its ad business, that number must come down.”
StackAdapt did not respond to ADWEEK’s request for comments.
A new decision layer
StackAdapt is justifying the spend with what it calls a “New Mid-funnel Decision Layer” in its deck, where ChatGPT ads are designed to influence users during research and comparison.
The environment functions more like a hybrid of search and display—surfacing brands against queries such as “best travel credit cards,” “top running shoes for beginners,” or “best CRM for small businesses,” the deck reads.
Ad formats are similar to what OpenAI currently runs on ChatGPT: a square (1:1) image paired with a 30-character headline and 60-character description, with tracking handled through static UTM parameters on landing URLs, labelled as ‘sponsored.’
Unlike traditional programmatic buying, the pilot strips away most familiar levers. There is no audience targeting, no demographic overlays and limited geographic controls, according to the agency executive.
Instead, everything hinges on the prompt. “We can do targeting based on prompt relevance, topic alignment or intent-based matching,” the executive added.
StackAdapt is leaning on ChatGPT’s scale to justify the spend citing 800 million weekly active users, more than 2 billion prompts, and claiming that roughly 10% of the world’s population now uses the platform.
Crucially for advertisers, those prompts often map to high-intent behaviors: people researching products, comparing options and seeking recommendations before making purchase decisions.
Once a user submits a query, OpenAI’s relevance model determines which ads are eligible to appear within the conversational response. There is limited information on how OpenAI’s relevance model works.
“Ads are selected by a separate model from the core chat model and do not influence ChatGPT’s answer,” the OpenAI spokesperson said. “Relevance is based on signals like the topic of the current conversation and, depending on a user’s settings, may also include past chats and prior ad interactions.”
A slide titled ‘Relevance First, Pricing Second’ outlines a filtering process that prioritizes how closely an ad’s headline, copy, landing page and keywords align with the user’s query. The most contextually aligned creative wins placement, even before auction dynamics fully kick in.
The StackAdapt program is a fully managed service, with limited reporting. Advertisers receive weekly impressions and clicks, with attribution handled through static UTM parameters.
“It really is just an innovation test,” the executive said.
The deck is clear about the experimental nature of the pilot. It warns of bugs, evolving auction mechanics and a rapidly changing ecosystem.
Participation also requires insertion orders with OpenAI-specific contractual language that is described as non-negotiable.
In StackAdapt’s framing, brands that show up early may have a hand in defining how advertising works inside AI systems. Whether that translates into measurable performance is still, by the company’s own admission, an open question.
Update 4/20/2026: This story was updated to include comment from OpenAI.