ChatGPT's First Advertisers Can't Prove Their Ads Work

A quarter of a million dollars. That's what one enterprise advertiser committed to OpenAI's ChatGPT ad pilot over several weeks. The amount actually spent? Just $7,500 — roughly 3% of the budget. The rest sat there, undelivered, while the advertiser waited for something to happen.

OpenAI's push into advertising was supposed to be the next logical step for a company burning through cash on AI infrastructure. Instead, early results paint a picture of a platform that isn't ready for prime time — and advertisers who signed up with high hopes are quietly admitting they can't prove the money was well spent.

The Numbers Tell the Story

The headline stat is brutal: ChatGPT ads are generating a click-through rate of 0.91%. For context, the Google Search benchmark in comparable sectors sits at 6.4% — nearly seven times higher. For agencies like WPP, Omnicom, and Dentsu, who committed minimum budgets of $200,000 to test the waters, that gap is hard to ignore.

But the CTR problem is only part of it. A major reporting glitch in OpenAI's newly launched Ads Manager has left advertisers unable to see their own campaign data. Instead of real-time dashboards, participants receive weekly CSV exports — a throwback that feels oddly analog for a company building the future of AI.

"Measurement capabilities appear to be troubled," said Ashley Fletcher, CMO at Adthena. "A major reporting glitch blocks visibility into their own data, making it all but impossible to optimize activity and understand ROI."

The Premium User Problem

There's a structural issue baked into ChatGPT's ad model that no amount of technical fixes can solve easily. The most engaged, highest-value ChatGPT users — the ones advertisers would most like to reach — pay $20 a month for Plus or Pro subscriptions. They will never see an ad.

Ads only appear on the free and Go tiers, served to logged-in adults in the U.S. That means the advertising inventory consists entirely of users who haven't been willing to pay for the product. As Alex Halliday of Ad Age put it: "ChatGPT just announced advertising. But the most valuable users in America will never see those ads. They're paying $20 a month for the privilege."

OpenAI partnered with Criteo for ad buying and targeting infrastructure, and the ads themselves appear at the bottom of ChatGPT answers when there's a relevant commercial connection to the conversation. The company has been careful to emphasize that ads don't influence AI responses and that conversation data won't be sold to advertisers. But good intentions don't fix a broken measurement pipeline.

Why OpenAI Needs This to Work

The motivation is straightforward: running GPT-5.4 and its variants across hundreds of millions of free users costs an enormous amount of money. OpenAI's infrastructure bill is growing faster than its subscription revenue, and advertising represents the most obvious path to monetizing the free tier without charging users directly.

The company first signaled interest in advertising back in December 2024. By December 2025, it had already pulled back on retail promotions within ChatGPT after users complained about what felt like hidden ads. The current pilot, launched in mid-March 2026, represents a more formal attempt — complete with a self-serve Ads Manager dashboard and plans for a full advertising ecosystem.

Initial spending proposals for newer advertisers range from $50,000 to $100,000, with the bigger agency commitments hitting $200,000 or more. OpenAI is clearly betting on volume, but so far the platform can't deliver on the basics that marketers expect.

What's Missing

"Building an ad ecosystem requires more than ad inventory," wrote Anu Adegbola at Search Engine Land. "Marketers expect robust reporting, optimization tools, and predictable performance." Right now, ChatGPT's ad platform offers none of the three.

The Ads Manager is still in its infancy. There's no A/B testing framework. Targeting options are limited. The weekly CSV reports are a far cry from the real-time optimization loops that advertisers have come to expect from Google and Meta. Fletcher summed it up diplomatically: "The platform is still finding its feet."

For comparison, Google spent over a decade refining its ad stack into the precision instrument it is today. Meta's advertising platform went through years of iteration before it became the revenue machine that now funds the company's AI ambitions. OpenAI is trying to build something comparable while simultaneously running the most expensive AI infrastructure in the world, and the growing pains are showing.

What Happens Next

OpenAI isn't backing down. The March 23 announcement confirmed that ads will roll out to all free and Go tier users in the U.S., signaling confidence that the pilot's problems are fixable. The company is reportedly working on enhanced analytics tools and promises that future updates will bring the kind of reporting and optimization capabilities advertisers need.

The bigger question is whether conversational AI advertising is fundamentally different from search advertising. When someone types a query into Google, they're often looking to buy something. When someone asks ChatGPT a question, they're looking for an answer. That intent gap may explain the CTR disparity more than any technical shortcoming — and it's not something a better dashboard can fix.

For now, the advertisers who signed up early are stuck in an awkward position: committed budgets, minimal spend, and no clear way to measure whether any of it mattered. The platform may mature. The reporting may improve. But the first chapter of ChatGPT advertising is being written in red ink.