• by Laurie Sullivan , Staff Writer, Yesterday

OpenAI’s ChatGPT last week began serving ads to non-logged in users, expanding the pool of available inventory and targeting options for brands using the platform.

But OpenAI's large language model (LLM) could develop an entirely new pricing structure beyond CPMs or cost-per-click (CPC) for advertising before the company begins to allow more advertisers to buy media.

Some advertisers in ChatGPT’s advertising pilot have not been satisfied with results from budgets that invested in the media, according to one outlet, which cites a digital ad executive. However, there has been proof the ads have served up to users.

The current basic digital advertising metrics like CPM for impression- or view-based media buys, or cost per click (CPC) that tells marketers how much they will pay each time someone clicks on their ad may not be the metrics that ChatGPT and other conversational bots will ultimately use.

The minimum spend seems to look “attractive on paper,” but many advertisers are having a difficult time running this through their budgets due to insufficient ad frequency and inventory.

OpenAI reportedly had to drop its ChatGPT ad-spend minimum from $200,000, when the pilot launched, to $50,000.

The source, who is not in the ad industry, said if she “hadn’t seen news stories about the introduction of ads in ChatGPT,” she might not have noticed them during her “10-to-15-minute conversation with the LLM about advice on how to structure her resume,” the report said. It also notes that “she was served two ads, one for Canva and one for JobCopilot.”

When I asked ChatGPT which brands have piloted ChatGPT Ads, it gave MediaPost a list, but upon more prompting, it told me the “list did not come from ads shown to me or ‘inside’ the platform UI. I don’t have visibility into what ads any user is being served. It came from external reporting and documentation.”

ChatGPT explained that the chatbot synthesizes what has been publicly reported about for the early ad pilot, not observing or pulling from live ads in ChatGPT itself.

“If you’re asking from a verification angle: there isn’t currently a way (even for me) to see a canonical, real-time advertiser list from inside the product,” ChatGPT told MediaPost.

When asked “why not,” ChatGPT explained that it is not connected to the live internet, but rather generates responses from training and general knowledge.

Without confirmation from an OpenAI executive, that will need to change based on the company’s ambitions to support the advertising industry.

ChatGPT disagreed with me, saying: “I’d push back on the idea that it has to change,” because for an advertising business to grow, it doesn’t require an “aware” product of advertisers. Platform providers build separate reporting, planning and buying systems around it.

“The revenue engine lives in dashboards, APIs, and measurement systems, not inside the user-facing experience,” according to ChatGPT. “If OpenAI follows that playbook (and early signals suggest it will), then advertisers get transparency via ads manager–style tools (campaign reporting, targeting, attribution). Users get a clean, trusted assistant that isn’t obviously steering answers based on ad demand. And, the model (me) stays blind to campaigns and spend, which protects both neutrality and privacy.”

ChatGPT wrote that its advertising model will evolve to more structured ad formats and better contextual matching.

“So the real shift won’t be inside the conversation — it’ll be in the tooling and data layer around it, where advertisers operate,” ChatGPT told MediaPost.

The model also said, the real inflection point isn’t “where the tools live” — it’s whether conversational AI can produce a better signal than search intent, “which would justify an entirely new pricing model beyond CPC.”